People

630 people who authored the works in this archive — built from authorship, not from who has a GitHub profile. A GitHub link is an optional badge, never a condition for being here, so foundational figures without a public developer identity are present alongside contemporary open-source builders. Contributor roles distinguish scientific originators from implementers, executives, and others. Co-authorship is listed as-is: it is not a claim of equal influence, and some grouped records under-list their authors.

630 people

Geoffrey HintonScientific originatorCo-authored backpropagation and developed dropout and layer normalization, the training methods deep networks are still trained with.9 works · from 1986 · ✓ verified
Ilya SutskeverScientific originatorCo-authored AlexNet and sequence-to-sequence learning, and drove the scaling of the GPT decoder-only line at OpenAI.7 works · from 2012 · ✓ verified
Noam ShazeerArchitecture researcherCo-invented the Transformer and later SwiGLU, multi-query attention, and sparse mixture-of-experts routing.7 works · from 2017 · ✓ verified
Quoc V. LeArchitecture researcherContributed across Google language-model work spanning sequence-to-sequence learning and instruction finetuning that let models follow tasks zero-shot.7 works · from 2014 · ✓ verified
Yoshua BengioScientific originatorBuilt the first neural language model and the attention mechanism for translation that later became the core of the Transformer.7 works · from 1994 · ✓ verified
Jason WeiScientific originatorIntroduced chain-of-thought prompting, instruction tuning (FLAN), and the emergent-abilities framing that shaped reasoning research.6 works · from 2021 · ✓ verified
Karthik NarasimhanScientific originatorCo-authored the original GPT and the ReAct and Reflexion agent frameworks combining reasoning with tool use.6 works · from 2018 · ✓ verified
Chris OlahSafety researcherFounded mechanistic interpretability, developing the circuits and induction-head framework for reverse-engineering how models compute.5 works · from 2020 · ✓ verified
Alec RadfordScientific originatorBuilt the original GPT and GPT-2 decoder-only models and CLIP, defining generative pretraining and vision-language contrastive learning.4 works · from 2018 · ✓ verified
Christopher ManningScientific originatorCo-created GloVe embeddings and Luong attention, and later co-authored Direct Preference Optimization for alignment.4 works · from 2014 · ✓ verified
Dario AmodeiExecutiveCo-authored RLHF and scaling work and leads Anthropic, one of the frontier labs shaping model capability and safety.4 works · from 2017 · ✓ verified
Jie TangArchitecture researcherLed Tsinghua's GLM family of open bilingual models, providing a widely used non-Western open-weight foundation.4 works · from 2021 · ✓ verified
John SchulmanScientific originatorDeveloped core policy-optimization methods for RLHF and studied reward overoptimization central to aligning language models.4 works · from 2021 · ✓ verified
Mike LewisArchitecture researcherCo-created BART's denoising encoder-decoder pretraining, unifying generation and understanding objectives.4 works · from 2020 · ✓ verified
Shunyu YaoScientific originatorCreated ReAct, Tree of Thoughts, and Reflexion, the reasoning-and-acting patterns that underpin modern LLM agents.4 works · from 2023 · ✓ verified
Alan TuringScientific originatorDefined computation with the Turing machine and framed machine intelligence, setting the conceptual ground for the whole field.3 works · from 1936 · probable
Catherine OlssonSafety researcherCo-discovered induction heads, a concrete mechanism explaining how Transformers perform in-context learning.3 works · from 2021 · ✓ verified
Christopher RéScientific originatorCo-developed FlashAttention and the S4 state-space models, driving efficient long-sequence architectures.3 works · from 2022 · ✓ verified
Claude ShannonScientific originatorFounded information theory, giving the mathematics of entropy and channel capacity that underlie how models compress and predict.3 works · from 1937 · probable
Demis HassabisExecutiveCo-founded DeepMind and led AlphaGo and AlphaFold, showing deep learning could solve protein structure and superhuman games.3 works · from 2016 · ✓ verified
Denny ZhouTraining-method researcherCo-created chain-of-thought and least-to-most prompting, methods that elicit multi-step reasoning from language models.3 works · from 2022 · ✓ verified
Ethan PerezSafety researcherBuilt red-teaming and evaluation methods for finding model failures and studied reward overoptimization in alignment.3 works · from 2020 · ✓ verified
Ion StoicaSystems researcherCo-led vLLM and PagedAttention, raising LLM serving throughput through paged KV-cache memory management.3 works · from 2023 · ✓ verified
Jan LeikeSafety researcherCo-developed learning from human preferences and led alignment teams building the oversight methods used to fine-tune deployed models.3 works · from 2017 · ✓ verified
John von NeumannScientific originatorDefined the stored-program computer architecture and co-developed Monte Carlo methods and game theory used throughout computing and AI.3 works · from 1944 · probable
Luke ZettlemoyerArchitecture researcherCo-created BART, a denoising encoder-decoder pretraining objective effective for text generation and comprehension.3 works · from 2020 · ✓ verified
Mark ChenScientific originatorLed Codex and HumanEval, launching code generation as a capability and benchmark, and directs research at OpenAI.3 works · from 2021 · probable
Neel NandaSafety researcherAdvanced and widely taught mechanistic interpretability methods, growing the field of reverse-engineering model computations.3 works · from 2021 · ✓ verified
Nelson ElhageSafety researcherCo-authored the mathematical framework for Transformer circuits and superposition, core tools for interpreting model internals.3 works · from 2021 · ✓ verified
Percy LiangBenchmark creatorLed HELM holistic evaluation and BIG-bench, standardizing how language models are measured across many tasks.3 works · from 2022 · ✓ verified
Ruslan SalakhutdinovTraining-method researcherCo-invented dropout regularization, reducing overfitting and enabling reliable training of larger neural networks.3 works · from 2006 · ✓ verified
Stefano ErmonTraining-method researcherCo-created Direct Preference Optimization, aligning models from preference data without a separate reward model or RL loop.3 works · from 2022 · ✓ verified
Tom B. BrownOpen-source implementerLead author of GPT-3, demonstrating few-shot in-context learning that reframed how large models are used.3 works · from 2017 · ✓ verified
Tri DaoSystems researcherCreated FlashAttention and co-invented Mamba, making attention IO-efficient and advancing state-space alternatives to it.3 works · from 2022 · ✓ verified
Weizhu ChenArchitecture researcherLed DeBERTa, improving BERT with disentangled attention and an enhanced mask decoder.3 works · from 2021 · ✓ verified
Albert GuArchitecture researcherCreated S4 and Mamba, the selective state-space models offering a linear-time alternative to attention.2 works · from 2022 · ✓ verified
Albert Q. JiangArchitecture researcherLed Mistral 7B and Mixtral, showing open sparse-MoE models could match far larger dense ones.2 works · from 2023 · ✓ verified
Alex GravesArchitecture researcherDeveloped CTC and Neural Turing Machines, connecting recurrent networks to external memory and end-to-end sequence labeling.2 works · from 2006 · ✓ verified
Alex KrizhevskyArchitecture researcherImplemented AlexNet, whose 2012 ImageNet win started the GPU deep-learning era.2 works · from 2012 · ✓ verified
Alexandre SablayrollesArchitecture researcherCo-authored Mixtral, a sparse mixture-of-experts open-weight model that matched larger dense models at lower inference cost.2 works · from 2023 · ✓ verified
Antoine BordesArchitecture researcherCo-authored the deep sparse rectifier work establishing ReLU activations as a default for training deep networks.2 works · from 2010 · ✓ verified
Aohan ZengArchitecture researcherCo-developed the GLM open model family and its blank-infilling pretraining approach.2 works · from 2021 · ✓ verified
Arthur MenschTheoretical researcherCo-authored the Chinchilla compute-optimal scaling result, then founded Mistral to ship open-weight models.2 works · from 2022 · ✓ verified
Ashish VaswaniArchitecture researcherLead author of 'Attention Is All You Need,' introducing the Transformer that underlies nearly every modern language model.2 works · from 2017 · ✓ verified
Barret ZophArchitecture researcherCo-created the Switch Transformer and ST-MoE, making sparse mixture-of-experts a practical way to scale parameter count.2 works · from 2022 · ✓ verified
Caiming XiongArchitecture researcherCo-created BLIP vision-language pretraining for unified image-text understanding and generation.2 works · from 2022 · ✓ verified
Carlos E. JimenezBenchmark creatorBuilt SWE-bench, the repository-level software-engineering benchmark now central to evaluating coding agents.2 works · from 2024 · ✓ verified
Collin BurnsSafety researcherCo-authored scalable-oversight work on supervising models for tasks humans cannot easily evaluate directly.2 works · from 2018 · ✓ verified
Dale SchuurmansTheoretical researcherCo-authored chain-of-thought prompting, showing that intermediate reasoning steps elicit multi-step problem solving from large models.2 works · from 2022 · ✓ verified
Damai DaiArchitecture researcherDesigned DeepSeekMoE's fine-grained expert specialization, a routing scheme adopted in efficient large open models.2 works · from 2024 · ✓ verified
Dan HendrycksBenchmark creatorBuilt MMLU and MATH, the knowledge and reasoning benchmarks used to compare frontier models, and works on AI safety.2 works · from 2021 · ✓ verified
Danqi ChenTraining-method researcherContributed dense-retrieval and preference-optimization methods that improved how models are grounded and aligned.2 works · from 2019 · ✓ verified
David SilverScientific originatorLed DQN and AlphaGo/AlphaZero, establishing deep reinforcement learning and self-play as methods for superhuman decision-making.2 works · from 2015 · ✓ verified
Daya GuoTraining-method researcherLead author of DeepSeek-R1, showing reinforcement learning alone can elicit strong reasoning in language models.2 works · from 2023 · ✓ verified
Diederik P. KingmaArchitecture researcherCo-invented the variational autoencoder, a foundational framework for deep latent-variable generative models.2 works · from 2013 · ✓ verified
Dzmitry BahdanauArchitecture researcherIntroduced the additive attention mechanism for translation that the Transformer generalized.2 works · from 2014 · ✓ verified
Ed H. ChiTheoretical researcherCo-authored the chain-of-thought prompting work showing that intermediate reasoning steps elicit better multi-step reasoning from LLMs.2 works · from 2022 · ✓ verified
Furu WeiArchitecture researcherLed multimodal and retrieval model efforts at Microsoft, including production vision-language models.2 works · from 2019 · ✓ verified
Gabriel GohSafety researcherCo-authored circuits-style interpretability work visualizing and naming features inside vision and multimodal models.2 works · from 2020 · ✓ verified
Gautier IzacardArchitecture researcherDeveloped Fusion-in-Decoder and the Atlas retrieval-augmented model, coupling retrieval tightly with generation.2 works · from 2021 · ✓ verified
Graham NeubigSystems researcherCo-created SWE-agent and OpenHands, frameworks and evaluations for autonomous software-engineering agents.2 works · from 2023 · ✓ verified
Guilherme PenedoDataset creatorBuilt the RefinedWeb dataset, showing carefully filtered web text alone can train strong open models.2 works · from 2023 · ✓ verified
Hieu PhamArchitecture researcherCo-authored effective attention-based neural machine translation approaches that refined how attention is applied in sequence models.2 works · from 2015 · ✓ verified
Hugo TouvronArchitecture researcherLed the LLaMA models whose open weights seeded much of the open large-language-model ecosystem.2 works · from 2023 · ✓ verified
HyoukJoong LeeSystems researcherCo-authored GShard, enabling automatic sharding to train giant mixture-of-experts models across accelerators.2 works · from 2018 · ✓ verified
Jacob HiltonSafety researcherStudied reward-model overoptimization and sycophancy, characterizing failure modes of RLHF-trained models.2 works · from 2021 · ✓ verified
Jakob UszkoreitArchitecture researcherCo-authored the Transformer and pushed the self-attention-only design that removed recurrence from sequence models.2 works · from 2017 · ✓ verified
Jason WestonArchitecture researcherCo-authored NLP-from-scratch and memory-network models, early steps toward unified neural architectures for language.2 works · from 2011 · ✓ verified
Jeff DeanExecutiveBuilt Google's core distributed-systems and MoE infrastructure and led its AI organization, shaping the scale at which models are trained.2 works · from 2015 · ✓ verified
Jeff WuTraining-method researcherCo-authored learning to summarize from human feedback, an early demonstration of RLHF on a real task.2 works · from 2020 · ✓ verified
Jian SunArchitecture researcherCo-invented ResNet residual connections, making it possible to train networks hundreds of layers deep.2 works · from 2010 · ✓ verified
Jingren ZhouExecutiveLed Alibaba's Qwen and vision-language model programs, shipping a widely used open-weight family.2 works · from 2023 · ✓ verified
Jinze BaiArchitecture researcherLed the Qwen and Qwen-VL open vision-language models, giving the community strong multilingual production-grade multimodal weights.2 works · from 2023 · ✓ verified
John YangBenchmark creatorCo-created SWE-bench and SWE-agent, defining both the task and a leading agent for autonomous software engineering.2 works · from 2024 · ✓ verified
Jong Wook KimOpen-source implementerCo-built CLIP and Whisper, the contrastive vision-language and robust speech-recognition models widely reused in multimodal systems.2 works · from 2021 · probable
Jürgen SchmidhuberArchitecture researcherCo-invented the LSTM and CTC, foundational recurrent architectures for speech and sequence learning.2 works · from 1997 · ✓ verified
Kaiming HeArchitecture researcherInvented residual connections (ResNet) and He initialization, which made training very deep networks stable.2 works · from 2010 · ✓ verified
Karl CobbeTraining-method researcherBuilt GSM8K and process-supervision verifiers, establishing step-level reward signals for training model reasoning.2 works · from 2021 · ✓ verified
Katherine LeeDataset creatorShowed training-data deduplication improves language models and quantified verbatim memorization risks.2 works · from 2020 · ✓ verified
Kenton LeeArchitecture researcherCo-authored BERT, the bidirectional masked-language-model pretraining that became the standard NLU baseline.2 works · from 2018 · ✓ verified
Kyunghyun ChoArchitecture researcherCo-invented the GRU and the RNN encoder-decoder with attention that reframed machine translation as neural sequence modeling.2 works · from 2014 · ✓ verified
Leo GaoDataset creatorLed The Pile, an open large-scale text corpus that enabled reproducible training of open language models.2 works · from 2021 · ✓ verified
Leon BottouTheoretical researcherCo-authored LeNet convolutional networks and foundational stochastic gradient descent methods for large-scale learning.2 works · from 1998 · ✓ verified
Lianmin ZhengBenchmark creatorBuilt Chatbot Arena and LLM-as-judge evaluation, the human-preference leaderboard used to rank chat models.2 works · from 2023 · ✓ verified
Long OuyangTraining-method researcherLead author of InstructGPT, the RLHF instruction-tuning recipe that turned raw language models into helpful assistants.2 works · from 2020 · ✓ verified
Maarten BosmaTraining-method researcherCo-authored FLAN instruction finetuning, showing that finetuning on many tasks yields zero-shot generalization.2 works · from 2022 · ✓ verified
Marvin MinskyScientific originatorCo-founded the field of AI at Dartmouth and, with Papert, framed the limits of single-layer perceptrons.2 works · from 1956 · ✓ verified
Matei ZahariaSystems researcherCo-created ColBERT late-interaction retrieval for efficient neural search and co-founded Databricks around scalable data systems.2 works · from 2020 · ✓ verified
Ming-Wei ChangArchitecture researcherCo-authored BERT and REALM, connecting bidirectional pretraining with retrieval-augmented language modeling.2 works · from 2018 · ✓ verified
Minh-Thang LuongArchitecture researcherIntroduced global and local attention mechanisms for neural machine translation, refining early attention.2 works · from 2015 · ✓ verified
Naman GoyalTraining-method researcherCo-authored RoBERTa, showing careful data and hyperparameter choices substantially improve BERT pretraining.2 works · from 2019 · ✓ verified
Nan DuArchitecture researcherLead author of GLaM, a sparse mixture-of-experts model that scaled capacity at reduced compute per token.2 works · from 2022 · ✓ verified
Niki ParmarArchitecture researcherCo-authored the Transformer, whose self-attention design replaced recurrence across sequence modeling.2 works · from 2017 · ✓ verified
Noah A. SmithArchitecture researcherCo-authored ALiBi position biasing, letting transformers extrapolate to sequences longer than those seen in training.2 works · from 2021 · ✓ verified
Ofir PressArchitecture researcherIntroduced ALiBi attention biasing, enabling Transformers to extrapolate to longer sequences than trained on.2 works · from 2021 · ✓ verified
Opher LieberArchitecture researcherCo-authored AI21's MRKL modular systems that combine language models with external tools and knowledge.2 works · from 2023 · ✓ verified
Oriol VinyalsScientific originatorCo-authored sequence-to-sequence learning and knowledge distillation, and led major reinforcement-learning systems at DeepMind.2 works · from 2014 · ✓ verified
Paul ChristianoSafety researcherOriginated reinforcement learning from human preferences and scalable-oversight methods that underlie modern alignment training.2 works · from 2017 · ✓ verified
Paulius MicikeviciusSystems researcherDeveloped mixed-precision training, which roughly doubled throughput and became standard for training large models.2 works · from 2018 · ✓ verified
Radu SoricutArchitecture researcherCo-authored ALBERT and Google's Conceptual Captions dataset for image-text pretraining.2 works · from 2019 · ✓ verified
Richard S. SuttonScientific originatorFounded temporal-difference learning, a cornerstone of modern reinforcement learning.2 works · from 1988 · ✓ verified
Richard SocherArchitecture researcherCo-created GloVe word embeddings from co-occurrence statistics, a widely used pre-transformer representation of word meaning.2 works · from 2009 · ✓ verified
Rico SennrichTraining-method researcherIntroduced byte-pair-encoding subword tokenization, the segmentation scheme most language models still use, and co-developed RMSNorm.2 works · from 2015 · ✓ verified
Rishi BommasaniResearch communicatorCo-authored the foundation-models framing and emergent-abilities analysis that shaped how the field describes scaled models.2 works · from 2022 · ✓ verified
Ronald J. WilliamsScientific originatorIntroduced REINFORCE, the policy-gradient method underlying much of reinforcement learning for sequences.2 works · from 1986 · ✓ verified
Sebastian BorgeaudTheoretical researcherCo-authored Chinchilla scaling laws and RETRO retrieval-augmented models at DeepMind.2 works · from 2021 · ✓ verified
Sepp HochreiterArchitecture researcherIdentified the vanishing-gradient problem and co-invented the LSTM that dominated sequence modeling before Transformers.2 works · from 1994 · ✓ verified
Shaoqing RenArchitecture researcherCo-invented ResNet and Faster R-CNN, foundational architectures for deep image recognition and detection.2 works · from 2010 · ✓ verified
Sharan NarangArchitecture researcherCo-authored T5, unifying diverse NLP tasks under a single text-to-text transfer-learning framework.2 works · from 2018 · ✓ verified
Stella BidermanDataset creatorCo-built The Pile and led open model suites, making large-scale training data and models publicly reproducible.2 works · from 2021 · ✓ verified
Taku KudoTraining-method researcherCreated SentencePiece and unigram tokenization, the language-agnostic subword tokenizers used across modern models.2 works · from 2018 · ✓ verified
Tim DettmersSystems researcherDeveloped 8-bit quantization and QLoRA, letting large models run and be fine-tuned on modest hardware.2 works · from 2022 · ✓ verified
Tom HenighanTheoretical researcherCo-authored the neural scaling laws quantifying loss as a power law in model size, data, and compute.2 works · from 2020 · ✓ verified
Wenfeng LiangExecutiveFounded DeepSeek and led its open-weight models, including the DeepSeek-R1 reasoning model.2 works · from 2024 · ✓ verified
William FedusArchitecture researcherCo-authored the Switch Transformer that simplified expert routing and pushed models past a trillion parameters.2 works · from 2021 · ✓ verified
Xavier GlorotTraining-method researcherIntroduced Xavier initialization and helped popularize ReLU activations, both standard for training deep networks.2 works · from 2010 · ✓ verified
Xiangyu ZhangArchitecture researcherCo-invented ResNet and efficient vision architectures used broadly across deep computer vision.2 works · from 2010 · ✓ verified
Xiao LiuArchitecture researcherWorked on the GLM family of pretrained language models and on agent/web benchmarks for evaluating LLM task execution.2 works · from 2021 · ✓ verified
Xuezhi WangTraining-method researcherCo-authored chain-of-thought prompting and self-consistency decoding, methods that boost multi-step reasoning.2 works · from 2022 · ✓ verified
Yanping HuangInfrastructure builderCo-developed GPipe pipeline parallelism, enabling training of models too large for a single accelerator.2 works · from 2019 · ✓ verified
Yanqi ZhouArchitecture researcherDeveloped expert-choice routing for mixture-of-experts models, improving load balance across experts.2 works · from 2020 · ✓ verified
Yi TayArchitecture researcherLed UL2's unified pretraining paradigm and surveyed efficient transformer architectures.2 works · from 2022 · ✓ verified
Ying ShengSystems researcherCreated FlexGen and SGLang, high-throughput systems for running and serving large language models efficiently.2 works · from 2023 · ✓ verified
Yinhan LiuTraining-method researcherLead author of RoBERTa, demonstrating BERT was undertrained and that recipe changes alone yield large gains.2 works · from 2019 · ✓ verified
Yoav ShohamExecutiveCo-founded AI21 and co-authored the MRKL modular tool-use architecture for augmenting language models.2 works · from 2023 · ✓ verified
Youlong ChengSystems researcherCo-authored GPipe, using pipeline parallelism to train very large models across many accelerators.2 works · from 2018 · ✓ verified
Yuanzhong XuSystems researcherCo-authored GShard for automatic sharding, a compiler-level path to training giant sparse models.2 works · from 2018 · ✓ verified
Zhilin YangArchitecture researcherCo-created XLNet and Transformer-XL, advancing autoregressive pretraining and longer-context attention.2 works · from 2018 · ✓ verified
Aarohi SrivastavaBenchmark creatorLead-organized BIG-bench, a large collaborative benchmark probing capabilities beyond standard tasks.1 work · from 2022 · ✓ verified
Aaron HarlapSystems researcherCo-authored PipeDream, introducing pipeline-parallel training with weight stashing for higher hardware utilization.1 work · from 2019 · probable
Abdelrahman MohamedArchitecture researcherCo-authored BART and early deep-learning speech-recognition pretraining methods.1 work · from 2020 · probable
Abhimanyu DubeyArchitecture researcherLed the Llama 3 model and co-authored Llama 2, major open-weight LLM releases from Meta.1 work · from 2024 · probable
Ada LovelaceScientific originatorWrote the first algorithm intended for a machine and foresaw computers manipulating symbols beyond numbers.1 work · from 1843 · probable
Adam RobertsArchitecture researcherCo-authored T5 and its text-to-text framework unifying NLP tasks under one training format.1 work · from 2020 · probable
Aditya RameshArchitecture researcherLed DALL-E, demonstrating high-quality text-to-image generation from a single generative model.1 work · from 2021 · ✓ verified
Adly TempletonSafety researcherCo-authored dictionary-learning interpretability that extracts monosemantic features from a model's activations.1 work · from 2023 · probable
Adrian WellerArchitecture researcherCo-authored linear-attention methods that reduce transformer attention cost from quadratic to linear.1 work · from 2020 · probable
Ahmed MurtadhaArchitecture researcherCo-authored RoFormer, introducing rotary position embeddings now standard in modern LLMs.1 work · from 2021 · probable
Aidan N. GomezArchitecture researcherCo-authored the Transformer paper that defined the dominant modern neural architecture.1 work · from 2017 · ✓ verified
Aitor LewkowyczTraining-method researcherCo-created Minerva, adapting a language model to solve quantitative and mathematical reasoning problems.1 work · from 2022 · ✓ verified
Aixin LiuArchitecture researcherCo-authored DeepSeek open-weight language models, part of a competitive open model lineage.1 work · from 2024 · probable
Aja HuangOpen-source implementerCore engineer on AlphaGo, the first program to defeat a professional Go player at full-board play.1 work · from 2016 · probable
Aleksandra PiktusArchitecture researcherCo-authored Retrieval-Augmented Generation, combining a parametric model with retrieved passages for knowledge-intensive tasks.1 work · from 2020 · probable
Alex AndonianSafety researcherWorked on knowledge editing, methods for locating and editing factual associations stored in a model's weights.1 work · from 2022 · probable
Alexander KirillovArchitecture researcherLed the Segment Anything Model and its data engine for promptable image segmentation at scale.1 work · from 2023 · ✓ verified
Alexander KolesnikovArchitecture researcherCo-created the Vision Transformer, showing pure transformers can match CNNs on large-scale image classification.1 work · from 2021 · probable
Alexander M. RushTheoretical researcherCo-authored data-constrained scaling analysis and widely used educational transformer implementations.1 work · from 2023 · probable
Alexander PritzelArchitecture researcherCo-authored AlphaFold, achieving highly accurate protein structure prediction from sequence.1 work · from 2021 · probable
Alexandra BirchArchitecture researcherCo-authored subword (BPE) tokenization, enabling open-vocabulary neural machine translation.1 work · from 2015 · probable
Alexey DosovitskiyArchitecture researcherCreated the Vision Transformer, showing image patches fed to a Transformer can match or beat convolutional networks.1 work · from 2021 · ✓ verified
Ali BehrouzArchitecture researcherDeveloped the Titans architecture with a neural test-time memory module for continual, long-horizon learning.1 work · from 2025 · ✓ verified
Alon HalevyDataset creatorArgued that scale of data, more than algorithmic sophistication, drives performance in learning systems.1 work · from 2009 · ✓ verified
Amar PhanishayeeSystems researcherCo-authored PipeDream, a pipeline-parallel training system for large deep models.1 work · from 2019 · probable
Amr AhmedArchitecture researcherCo-authored sparse-attention transformers (BigBird) that scale attention to much longer sequences.1 work · from 2020 · probable
An YangArchitecture researcherCo-authored Alibaba's Qwen open-weight model family.1 work · from 2024 · probable
Anders AndreassenArchitecture researcherCo-authored Minerva, a language model finetuned to solve quantitative and mathematical reasoning problems.1 work · from 2022 · probable
Andrew GuSystems researcherCo-authored PyTorch FSDP, bringing sharded data-parallel training into the standard framework.1 work · from 2023 · probable
Andrew M. DaiArchitecture researcherCo-authored GLaM, an efficient mixture-of-experts language model activating only part of its parameters per token.1 work · from 2022 · probable
Andrew NystromDataset creatorCo-authored work showing that deduplicating training data measurably improves language models.1 work · from 2022 · probable
Andrey KolmogorovTheoretical researcherCo-founded algorithmic information theory, formalizing the complexity and compressibility of data central to learning theory.1 work · from 1964 · probable
Andy DavisArchitecture researcherCo-authored the sparsely-gated mixture-of-experts layer that introduced conditional computation at scale.1 work · from 2017 · probable
Angela FanArchitecture researcherContributor to BLOOM, the open multilingual large language model from the BigScience effort.1 work · from 2022 · probable
Angelina McMillan-MajorGovernance researcherCo-authored Stochastic Parrots, articulating social and environmental risks of ever-larger language models.1 work · from 2021 · probable
Angelos KatharopoulosArchitecture researcherCo-authored linear-attention transformers framed as RNNs, enabling fast autoregressive inference.1 work · from 2020 · probable
Anima AnandkumarArchitecture researcherCo-authored Voyager, an LLM-driven agent that autonomously acquires skills in an open-ended environment.1 work · from 2023 · probable
Ankur BapnaSystems researcherCo-authored GPipe, applying pipeline parallelism to train large models across accelerators.1 work · from 2019 · probable
Anthony BrohanArchitecture researcherCo-authored SayCan, grounding language-model planning in a robot's real-world affordances.1 work · from 2022 · probable
Antoine RouxArchitecture researcherCo-authored Mixtral of Experts, a sparse mixture-of-experts language model.1 work · from 2024 · probable
Archit SharmaTraining-method researcherCo-authored Direct Preference Optimization (DPO), aligning language models from preference data without a separate reward model.1 work · from 2023 · probable
Arman CohanArchitecture researcherWorked on sparse attention methods for extending Transformer context efficiently.1 work · from 2020 · probable
Ashwin GopinathOpen-source implementerCo-authored Reflexion, an approach where language agents improve by verbally reflecting on feedback from prior attempts.1 work · from 2023 · probable
Atri RudraSystems researcherCo-authored FlashAttention, an IO-aware exact-attention algorithm that speeds up Transformer training and inference.1 work · from 2022 · probable
Augustus OdenaBenchmark creatorCo-authored the HumanEval and MBPP benchmarks for evaluating code generation by language models.1 work · from 2021 · probable
Azalia MirhoseiniArchitecture researcherCo-authored 'Outrageously Large Neural Networks', introducing the sparsely-gated mixture-of-experts layer.1 work · from 2017 · probable
Baptiste RozièreArchitecture researcherWorked on open code language models (Code Llama) for program synthesis and infilling.1 work · from 2023 · probable
Barak LenzInfrastructure builderWorked on function calling, MRKL and the Model Context Protocol, connecting language models to external tools and knowledge.1 work · from 2023 · probable
Barlas OguzArchitecture researcherCo-authored Dense Passage Retrieval, a learned dual-encoder retriever for open-domain question answering.1 work · from 2020 · probable
Barry HaddowDataset creatorCo-authored the subword/BPE tokenization method for handling rare words in neural machine translation.1 work · from 2015 · probable
Bart van MerriënboerArchitecture researcherCo-authored the RNN encoder-decoder and gated recurrent unit (GRU) for sequence-to-sequence modeling.1 work · from 2014 · probable
Bei ChenTraining-method researcherDeveloped execution-feedback and unit-test-based methods that improve code generation reliability.1 work · from 2022 · ✓ verified
Benjamin MannArchitecture researcherContributor to GPT-3 ('Language Models are Few-Shot Learners'), demonstrating in-context few-shot learning at scale.1 work · from 2020 · probable
Bernard GhanemSystems researcherWorked on agent memory and multi-agent systems for coordinating multiple LLM-based agents.1 work · from 2023 · probable
Biao ZhangArchitecture researcherIntroduced RMSNorm, a simplified normalization now standard in most large language models.1 work · from 2019 · ✓ verified
Bill PeeblesArchitecture researcherCo-authored the Diffusion Transformer and video generation models used as world simulators.1 work · from 2024 · probable
Bill TutteScientific originatorBroke the German Lorenz cipher at Bletchley Park, foundational codebreaking work in early computing.1 work · from 1943 · probable
Bing XuSystems researcherCo-authored 'Training Deep Nets with Sublinear Memory Cost', introducing gradient checkpointing to cut training memory.1 work · from 2016 · probable
Binyuan HuiArchitecture researcherWorked on open code language models (Qwen Coder) for program generation.1 work · from 2023 · probable
Björn OmmerArchitecture researcherLed latent diffusion models (Stable Diffusion) and diffusion transformers for image generation.1 work · from 2022 · probable
Blake HechtmanInfrastructure builderWorked on Mesh-TensorFlow and GSPMD for automatic sharding and parallelizing large model training.1 work · from 2018 · probable
Bo PengArchitecture researcherCreated RWKV, an RNN-transformer hybrid delivering transformer-quality with linear-time inference for long context.1 work · from 2023 · probable
Bo WenArchitecture researcherCo-authored RoFormer, introducing rotary position embeddings (RoPE) now standard in Transformers.1 work · from 2021 · probable
Boaz BarakTheoretical researcherCo-authored 'Scaling Data-Constrained Language Models', studying how to scale training under limited data.1 work · from 2023 · probable
Boris GinsburgBenchmark creatorWorked on evaluating long-context reliability, showing models struggle to use information placed in the middle of long inputs.1 work · from 2023 · probable
Bowen PengArchitecture researcherWorked on context-length extension methods (YaRN) that rescale rotary embeddings to lengthen usable context.1 work · from 2023 · probable
Brando MirandaTheoretical researcherCo-authored 'Are Emergent Abilities of LLMs a Mirage?', arguing apparent emergence can be an artifact of the chosen metric.1 work · from 2023 · probable
Brian IchterTheoretical researcherCo-authored the chain-of-thought prompting work eliciting multi-step reasoning from language models.1 work · from 2022 · probable
Cade GordonDataset creatorCo-created LAION-5B, the large open image-text dataset for training vision-language models.1 work · from 2022 · probable
Carlos GuestrinSystems researcherCo-authored 'Training Deep Nets with Sublinear Memory Cost', trading compute for memory via activation checkpointing.1 work · from 2016 · probable
Chao JiaArchitecture researcherLead author of ALIGN, scaling contrastive image-text pretraining on noisy web data without curation.1 work · from 2021 · ✓ verified
Charles BabbageScientific originatorDesigned the Analytical Engine, the first general-purpose mechanical computer concept.1 work · from 1843 · probable
Charles PackerArchitecture researcherCreated MemGPT, giving agents operating-system-style management of long-term memory beyond the context window.1 work · from 2023 · ✓ verified
Charles SuttonBenchmark creatorCo-authored the HumanEval and MBPP code-generation benchmarks.1 work · from 2021 · probable
Charlie ChenSystems researcherWorked on speculative decoding (and Medusa/EAGLE), accelerating LLM inference with draft-and-verify token generation.1 work · from 2023 · probable
Chelsea FinnTraining-method researcherCo-authored Direct Preference Optimization (DPO) for aligning models directly from preference data.1 work · from 2023 · probable
Cheng-Ping HsiehBenchmark creatorWorked on long-context reliability benchmarks (RULER) probing how well models use long inputs.1 work · from 2023 · probable
Chenggang ZhaoArchitecture researcherContributor to DeepSeekMoE, a fine-grained mixture-of-experts architecture.1 work · from 2024 · probable
Chengqi DengArchitecture researcherContributor to DeepSeekMoE, a fine-grained mixture-of-experts language model.1 work · from 2024 · probable
Chi WangSystems researcherCreated AutoGen for building multi-agent LLM systems with shared memory and coordination.1 work · from 2023 · probable
Chitwan SahariaArchitecture researcherLead author of Imagen, a text-to-image diffusion model showing large frozen text encoders drive image fidelity.1 work · from 2022 · ✓ verified
Chiyuan ZhangSystems researcherCo-authored 'Training Deep Nets with Sublinear Memory Cost' on memory-efficient backpropagation.1 work · from 2016 · probable
Chris HallacyArchitecture researcherCo-authored CLIP, learning joint image-text representations via contrastive pretraining.1 work · from 2021 · probable
Chris J. MaddisonTraining-method researcherCo-authored AlphaGo, combining deep networks with self-play reinforcement learning and tree search to master Go.1 work · from 2016 · probable
Christian JauvinScientific originatorCo-authored 'A Neural Probabilistic Language Model' (Bengio et al.), an early neural language model with learned word embeddings.1 work · from 2003 · probable
Christian SzegedyTraining-method researcherCo-authored Batch Normalization, a technique that stabilizes and speeds up deep network training.1 work · from 2015 · probable
Christoph SchuhmannDataset creatorCo-created LAION-5B, the large open image-text dataset used to train open vision-language and diffusion models.1 work · from 2022 · ✓ verified
Christopher AkikiArchitecture researcherContributor to BLOOM, the open multilingual large language model from the BigScience effort.1 work · from 2022 · probable
Christopher J. C. H. WatkinsTheoretical researcherInvented Q-learning, the off-policy value-iteration algorithm at the base of deep reinforcement learning.1 work · from 1992 · ✓ verified
Chunyuan LiArchitecture researcherCo-authored LLaVA, connecting a vision encoder to an LLM for visual instruction following.1 work · from 2023 · probable
Colin RaffelArchitecture researcherLed T5, casting every NLP task as text-to-text and setting a widely-used transfer-learning framework.1 work · from 2020 · ✓ verified
Colin WhiteBenchmark creatorWorked on contamination-resistant benchmarks that resist test-set leakage into training data.1 work · from 2021 · probable
Dan AlistarhSystems researcherWorked on post-training inference quantization (e.g. GPTQ) to compress LLMs for efficient serving.1 work · from 2022 · probable
Dan JurafskyArchitecture researcherWorked on ColBERT and late-interaction retrieval/reranking models for efficient neural search.1 work · from 2020 · probable
Daniel FriedArchitecture researcherWorked on code language models (InCoder, CodeGen) with fill-in-the-middle infilling.1 work · from 2022 · probable
Daniel HesslowDataset creatorCo-authored the RefinedWeb dataset used to train the Falcon LLMs, showing filtered web data can rival curated corpora.1 work · from 2023 · probable
Daniel KhashabiTraining-method researcherCo-authored FLAN ('Finetuned Language Models Are Zero-Shot Learners'), showing instruction tuning improves zero-shot generalization.1 work · from 2022 · probable
Daniel Y. FuSystems researcherCo-authored FlashAttention, the IO-aware attention algorithm for faster Transformers.1 work · from 2022 · probable
Daniel ZieglerSafety researcherCo-authored 'Learning to Summarize from Human Feedback', an early RLHF method aligning outputs to human preferences.1 work · from 2020 · probable
Danny HernandezTheoretical researcherCo-authored 'AI and Compute', quantifying the exponential growth of compute used to train frontier models.1 work · from 2018 · probable
Daphne IppolitoDataset creatorCo-authored work showing that deduplicating training data improves language models and reduces memorization.1 work · from 2022 · probable
Darren EdgeArchitecture researcherCreated GraphRAG, combining knowledge-graph structure with retrieval for global sensemaking over corpora.1 work · from 2024 · ✓ verified
David BauSafety researcherWorked on mechanistic interpretability and knowledge editing, locating and modifying facts stored in model weights.1 work · from 2022 · probable
David ChoiArchitecture researcherContributor to AlphaCode, a system for competitive-programming code generation.1 work · from 2022 · probable
David DohanArchitecture researcherCo-authored Minerva, a model adapted for quantitative and mathematical reasoning problems.1 work · from 2022 · probable
David LuanArchitecture researcherContributor to GPT-2 ('Language Models are Unsupervised Multitask Learners').1 work · from 2019 · probable
David ReinBenchmark creatorCreated GPQA, a graduate-level question-answering benchmark hard enough to resist casual web lookup.1 work · from 2023 · ✓ verified
David RumelhartScientific originatorCo-authored the 1986 backpropagation paper that made training multi-layer neural networks practical.1 work · from 1986 · ✓ verified
Deepak NarayananInfrastructure builderDeveloped PipeDream pipeline parallelism and Megatron-scale training systems for large models.1 work · from 2019 · ✓ verified
Di HeArchitecture researcherCo-authored analysis of layer normalization placement in Transformers, motivating the Pre-LN design for stable training.1 work · from 2020 · probable
Dian YuOpen-source implementerCo-authored ReAct, interleaving reasoning traces and actions so language models can plan and use tools as agents.1 work · from 2023 · probable
Dirk GroeneveldInfrastructure builderLed OLMo's fully open training stack, releasing model, data and code for reproducible LLM training.1 work · from 2024 · probable
Dmitry LepikhinSystems researcherBuilt GShard, the automatic-sharding system that made training giant mixture-of-experts models across many devices feasible.1 work · from 2021 · ✓ verified
Donald HebbScientific originatorProposed Hebbian learning, the principle that co-active neurons strengthen connections, foundational to learning rules.1 work · from 1949 · probable
Dong YuArchitecture researcherCo-authored early deep neural network acoustic models for speech recognition, a foundational application of deep learning.1 work · from 2012 · probable
Dongxu LiArchitecture researcherCo-authored BLIP and BLIP-2, vision-language models bridging frozen image encoders and LLMs.1 work · from 2023 · probable
Douwe KielaTraining-method researcherContributed to the preference-optimization family of alignment methods that tune language models directly on human preference comparisons.1 work · from 2024 · probable
Edouard GraveArchitecture researcherCo-developed retrieval-augmented language models (Fusion-in-Decoder, RETRO, Atlas) that ground generation in retrieved passages.1 work · from 2021 · probable
Edward BermanOpen-source implementerCo-authored Reflexion, the agent method that improves task performance by storing verbal self-reflection on feedback as memory.1 work · from 2023 · probable
Edward J. HuTraining-method researcherInvented LoRA, the low-rank adaptation method that made fine-tuning large models cheap and modular.1 work · from 2022 · ✓ verified
Ehud KarpasArchitecture researcherCo-designed MRKL systems, routing language models to external tools and modules for reliable computation.1 work · from 2023 · ✓ verified
Elias FrantarSystems researcherDeveloped GPTQ post-training quantization, making low-bit LLM inference practical without retraining.1 work · from 2022 · ✓ verified
Emily M. BenderGovernance researcherCo-authored 'Stochastic Parrots,' framing the data, bias, and scale risks that shaped responsible-AI debate.1 work · from 2021 · ✓ verified
Enrico ShippoleArchitecture researcherWorked on context-window extension methods that rescale positional encodings to lengthen the sequences transformers can process.1 work · from 2023 · probable
Eric MitchellTraining-method researcherCo-authored Direct Preference Optimization (DPO), aligning models from preference pairs without a separate reward model or RL loop.1 work · from 2023 · probable
Erik NijkampArchitecture researcherCo-created open code-generation models (CodeGen, InCoder), including fill-in-the-middle infilling for program synthesis.1 work · from 2022 · probable
Faustino GomezTraining-method researcherCo-authored Connectionist Temporal Classification, the alignment-free loss enabling neural sequence labeling for speech and handwriting.1 work · from 2006 · probable
Federico CassanoOpen-source implementerContributed to Reflexion, building the self-reflective agent loop that retries tasks using natural-language feedback.1 work · from 2023 · probable
Federico LebronArchitecture researcherContributed to Grouped-Query Attention (GQA), which shrinks the key/value cache to speed up transformer inference.1 work · from 2023 · probable
Fei XiaTraining-method researcherCo-authored Chain-of-Thought prompting, showing that generating intermediate reasoning steps improves complex problem-solving.1 work · from 2022 · probable
Fei-Fei LiDataset creatorCreated ImageNet, the large labeled dataset and benchmark whose scale enabled the deep-learning breakthrough in vision.1 work · from 2009 · probable
Felix GersArchitecture researcherCo-developed the LSTM forget gate, letting recurrent networks learn when to reset their own memory.1 work · from 1997 · probable
Fernando PereiraTheoretical researcherCo-authored 'The Unreasonable Effectiveness of Data,' arguing large corpora outweigh algorithmic cleverness for language tasks.1 work · from 2009 · probable
François FleuretArchitecture researcherCo-authored linear-attention transformers, reformulating attention as a kernel to achieve linear-time sequence processing.1 work · from 2020 · probable
Frank RosenblattScientific originatorBuilt the perceptron, the first trainable artificial neuron and the ancestor of all neural networks.1 work · from 1958 · probable
Gabriel SynnaeveArchitecture researcherLed open code models (Code Llama), adapting large language models for programming and infilling tasks.1 work · from 2023 · probable
Gemma Team (Google DeepMind)Architecture researcherProduced the open-weights Gemma model family, releasing DeepMind's lightweight models built on Gemini research to the ecosystem.1 work · from 2024 · probable
Geoffrey IrvingSafety researcherDeveloped scalable-oversight approaches like debate for supervising models on tasks humans cannot directly check.1 work · from 2018 · ✓ verified
Geon-Woo KimSystems researcherCo-authored Orca, the serving system that introduced iteration-level (continuous) batching for transformer inference.1 work · from 2022 · probable
George BooleTheoretical researcherCreated Boolean algebra, the logic of true/false operations on which all digital computation rests.1 work · from 1854 · probable
George CybenkoTheoretical researcherProved the universal approximation theorem, establishing that neural networks can represent any continuous function.1 work · from 1989 · ✓ verified
Georgi GerganovOpen-source implementerCreated llama.cpp and ggml, enabling efficient LLM inference on commodity CPUs and consumer hardware.1 work · from 2023 · ✓ verified
Gerald TesauroScientific originatorBuilt TD-Gammon, an early demonstration of temporal-difference reinforcement learning with neural networks.1 work · from 1995 · ✓ verified
Google DeepMindAuthor of Native/Omni Multimodality in the archive.1 work · from 2024 · probable
Gordon WelchmanScientific originatorDevised traffic analysis and the Bombe's diagonal board at Bletchley Park, industrializing machine-assisted Enigma codebreaking.1 work · from 1943 · probable
Greg BrockmanExecutiveCo-founded OpenAI and led the engineering behind its large-scale training and deployment infrastructure.1 work · from 2022 · probable
Greg CorradoArchitecture researcherCo-authored word2vec, whose efficient skip-gram and CBOW training made dense word embeddings ubiquitous.1 work · from 2013 · probable
Greg WayneArchitecture researcherCo-developed Neural Turing Machines and differentiable memory, coupling neural controllers to external addressable memory.1 work · from 2014 · probable
Grégoire MialonBenchmark creatorBuilt agent and web-navigation benchmarks (GAIA) for evaluating tool-using assistants on real-world tasks.1 work · from 2023 · probable
Gregory ChaitinScientific originatorFounded algorithmic information theory, defining complexity as the length of the shortest program that outputs a given string.1 work · from 1964 · probable
Guangxuan XiaoSystems researcherCo-authored SmoothQuant, shifting activation outliers into weights to enable accurate 8-bit LLM inference.1 work · from 2022 · probable
Guanzhi WangArchitecture researcherCreated Voyager, an LLM-driven agent that autonomously explores and acquires skills in an open-ended environment.1 work · from 2023 · ✓ verified
Guillaume LampleArchitecture researcherCo-founded Mistral and built Mistral 7B and the Mixtral mixture-of-experts, efficient open-weight models.1 work · from 2023 · probable
Guohao LiArchitecture researcherCreated CAMEL and related multi-agent frameworks for role-playing communication and memory among cooperating LLM agents.1 work · from 2023 · probable
Gwern BranwenResearch communicatorArticulated the scaling hypothesis in widely-read writing that shaped how the field interpreted larger models.1 work · from 2020 · probable
Gyeong-In YuSystems researcherDeveloped Orca's iteration-level continuous batching, a core technique for high-throughput LLM serving.1 work · from 2022 · ✓ verified
Ha TrinhArchitecture researcherWorked on graph-based structured retrieval (GraphRAG), organizing knowledge into graphs for retrieval-augmented generation.1 work · from 2024 · probable
Han HuInfrastructure builderCo-authored FP8-LM, demonstrating stable large-model training in 8-bit floating point to cut compute cost.1 work · from 2023 · probable
Hannaneh HajishirziTraining-method researcherContributed to instruction tuning (FLAN), finetuning models on many tasks phrased as instructions to unlock zero-shot generalization.1 work · from 2022 · probable
Hanxiao LiuArchitecture researcherCo-authored expert-choice routing for mixture-of-experts, letting experts select tokens to balance load across the network.1 work · from 2022 · probable
Hao LiuInfrastructure builderDeveloped Ring Attention, distributing attention across devices to extend context length far beyond single-device limits.1 work · from 2024 · ✓ verified
Haotian LiuArchitecture researcherBuilt LLaVA, the open visual-instruction-tuning recipe that made multimodal chat models easy to reproduce.1 work · from 2023 · probable
Harm de VriesDataset creatorHelped build open code models and datasets (StarCoder, The Stack) through the BigCode project.1 work · from 2023 · probable
Heewoo JunArchitecture researcherContributed to Codex, the GPT model fine-tuned on code that generates programs from natural-language prompts.1 work · from 2021 · probable
Hongyang ZhangSystems researcherCo-developed speculative-decoding methods (Medusa, EAGLE) that draft and verify tokens to accelerate LLM inference.1 work · from 2023 · probable
Houwen PengInfrastructure builderDeveloped FP8-LM, showing FP8 numerics can train large models with reduced memory and compute cost.1 work · from 2023 · ✓ verified
Hunter LightmanTraining-method researcherCo-authored 'Let's Verify Step by Step,' training process-supervised verifiers that reward correct reasoning steps, not just final answers.1 work · from 2023 · probable
Ian J. GoodfellowArchitecture researcherInvented Generative Adversarial Networks, the generative-modeling framework behind a decade of image synthesis work.1 work · from 2014 · ✓ verified
Ilia ShumailovDataset creatorIdentified model collapse, the degradation that occurs when models are trained on their own generated data.1 work · from 2024 · ✓ verified
Illia PolosukhinArchitecture researcherCo-authored the Transformer paper, one of the eight originators of the attention-based architecture.1 work · from 2017 · ✓ verified
Irwan BelloArchitecture researcherCo-authored ST-MoE, stabilizing the training of sparse mixture-of-experts transformers at scale.1 work · from 2022 · probable
Iryna GurevychArchitecture researcherCo-created Sentence-BERT and contrastive text-embedding methods for efficient semantic similarity and retrieval.1 work · from 2019 · probable
Ishan MisraArchitecture researcherCo-developed foundation vision models (Segment Anything, ImageBind) for promptable segmentation and multimodal embedding.1 work · from 2023 · probable
Ivo DanihelkaArchitecture researcherContributed to Neural Turing Machines and differentiable neural computers, adding external memory to neural controllers.1 work · from 2014 · probable
Iz BeltagyArchitecture researcherCreated Longformer's sparse attention pattern, scaling transformers to long documents in linear memory.1 work · from 2020 · ✓ verified
Izhak ShafranTraining-method researcherCo-authored ReAct, interleaving reasoning traces with tool actions so language models can plan and act.1 work · from 2023 · probable
J. Robert OppenheimerExecutiveDirected Los Alamos in the Manhattan Project, a founding model of large-scale coordinated scientific effort.1 work · from 1945 · probable
Jack W. RaeArchitecture researcherLed retrieval-augmented LM work (RETRO), retrieving from a trillion-token database to augment a transformer at inference.1 work · from 2021 · probable
Jacob AustinBenchmark creatorCo-created the MBPP program-synthesis benchmark for evaluating code generation from natural-language prompts.1 work · from 2021 · probable
Jacob DevlinArchitecture researcherLead author of BERT, the bidirectional encoder pretraining that became the default for language understanding tasks.1 work · from 2018 · ✓ verified
Jacob SteinhardtBenchmark creatorCo-created the MATH and related reasoning benchmarks that measure mathematical problem-solving in language models.1 work · from 2021 · probable
James KirkpatrickTraining-method researcherDeveloped elastic weight consolidation to reduce catastrophic forgetting during sequential task learning.1 work · from 2017 · ✓ verified
James Lee-ThorpArchitecture researcherContributed to Grouped-Query Attention (GQA), interpolating between multi-head and multi-query attention for efficient decoding.1 work · from 2023 · probable
James ThorneTraining-method researcherContributed to preference-optimization alignment methods for tuning language models on human preference data.1 work · from 2024 · probable
Jamie Ryan KirosTraining-method researcherCo-authored Layer Normalization, the per-sample normalization that stabilizes training of recurrent and transformer models.1 work · from 2016 · probable
Jane Dwivedi-YuTraining-method researcherCo-authored Toolformer, teaching language models to self-supervise when and how to call external APIs.1 work · from 2023 · probable
Jared KaplanTheoretical researcherDerived the neural scaling laws that predict model performance from compute, data, and parameters, guiding how labs allocate resources.1 work · from 2020 · ✓ verified
Jay ShahSystems researcherCo-authored FlashAttention-3, exploiting Hopper GPU features to push attention closer to hardware peak throughput.1 work · from 2022 · probable
Jean Pouget-AbadieArchitecture researcherCo-authored the original Generative Adversarial Networks paper, framing generation as a minimax game between generator and discriminator.1 work · from 2014 · probable
Jean-Baptiste AlayracArchitecture researcherBuilt Flamingo, an early visual language model doing few-shot learning across interleaved image and text.1 work · from 2022 · ✓ verified
Jeff DonahueArchitecture researcherCo-authored Flamingo, bridging frozen vision and language models for few-shot multimodal learning.1 work · from 2022 · probable
Jeff RasleyInfrastructure builderCo-authored ZeRO and DeepSpeed, partitioning optimizer state across devices to train models too large for one GPU.1 work · from 2020 · probable
Jeffrey DeanInfrastructure builderCo-authored word2vec and built the distributed systems (DistBelief, TensorFlow) that made large-scale neural training possible.1 work · from 2013 · probable
Jeffrey ElmanScientific originatorIntroduced simple recurrent networks, showing neural nets can learn temporal structure in sequences.1 work · from 1990 · probable
Jeffrey PenningtonArchitecture researcherCo-created GloVe, deriving word embeddings from global co-occurrence statistics.1 work · from 2014 · probable
Jeffrey QuesnelleArchitecture researcherCo-authored YaRN, extending transformer context length by rescaling rotary position embeddings.1 work · from 2023 · probable
Jeffrey WuArchitecture researcherContributed to GPT-2, showing that large unsupervised language models perform many tasks zero-shot.1 work · from 2019 · probable
Jeffrey ZhaoTraining-method researcherCo-authored ReAct, combining chain-of-thought reasoning with action steps for tool-using agents.1 work · from 2023 · probable
Jerry TworekArchitecture researcherContributed to Codex, OpenAI's code model behind program synthesis from natural language.1 work · from 2021 · probable
Ji LinSystems researcherCo-authored AWQ activation-aware weight quantization, preserving accuracy for low-bit LLM inference.1 work · from 2022 · probable
Jia DengDataset creatorCo-built ImageNet, the dataset that made large-scale supervised visual learning possible.1 work · from 2009 · probable
Jian-Guang LouTraining-method researcherWorked on execution-feedback methods that use unit tests and program outputs to improve code generation.1 work · from 2022 · probable
Jianfeng GaoArchitecture researcherCo-authored DeBERTa, whose disentangled attention and enhanced mask decoder improved encoder pretraining.1 work · from 2021 · probable
Jianlin SuArchitecture researcherInvented rotary position embeddings (RoPE), the positional encoding used by most current large language models.1 work · from 2021 · ✓ verified
Jifeng DaiArchitecture researcherLed open vision-language models (InternVL) and earlier deformable-convolution architectures for perception.1 work · from 2023 · probable
Jimmy BaTraining-method researcherCo-authored the Adam optimizer, the adaptive-moment gradient method that became the default for training deep networks.1 work · from 2014 · probable
Jimmy Lei BaArchitecture researcherCo-invented layer normalization, a stabilization technique central to transformer training.1 work · from 2016 · probable
Jingfei DuArchitecture researcherContributed to RoBERTa, showing that longer training on more data with tuned hyperparameters substantially improves BERT.1 work · from 2019 · probable
Jiwoo HongTraining-method researcherCo-authored ORPO, a reference-free preference-optimization method that folds alignment into supervised fine-tuning.1 work · from 2024 · probable
John JumperScientific originatorLed AlphaFold, whose accurate protein-structure prediction became a landmark scientific application of deep learning.1 work · from 2021 · probable
John McCarthyScientific originatorCoined 'artificial intelligence' and organized the 1956 Dartmouth workshop that founded the field.1 work · from 1956 · probable
John RichardsonDataset creatorCo-authored SentencePiece, the language-agnostic subword tokenizer used across modern LLMs.1 work · from 2018 · probable
Jonah AlbenSystems researcherCo-authored mixed-precision training, using FP16 with loss scaling to halve memory and speed training on tensor cores.1 work · from 2018 · probable
Joo Seong JeongSystems researcherCo-authored Orca, the transformer serving system introducing iteration-level scheduling for higher inference throughput.1 work · from 2022 · probable
Joon Sung ParkArchitecture researcherCreated Generative Agents, simulating believable human behavior with memory, reflection, and planning modules.1 work · from 2023 · probable
Jordan HoffmannScientific originatorLead author of Chinchilla, showing compute-optimal training needs far more data per parameter and resetting scaling practice.1 work · from 2022 · probable
Joseph E. GonzalezSystems researcherCo-developed systems for LLM agents (MemGPT, Gorilla) that add memory management and tool use atop serving infrastructure.1 work · from 2023 · probable
Joshua AinslieArchitecture researcherIntroduced grouped-query attention, cutting inference memory bandwidth while preserving quality.1 work · from 2023 · probable
Joshua BatsonSafety researcherCo-authored 'Towards Monosemanticity,' using sparse autoencoders to decompose transformer activations into interpretable features.1 work · from 2023 · probable
Julian MichaelSafety researcherWorked on reasoning-corrective and oversight methods for eliciting more truthful, checkable model reasoning.1 work · from 2023 · probable
Julian SchrittwieserArchitecture researcherCo-authored AlphaGo, AlphaZero, and MuZero, combining deep networks with tree search to master games from self-play.1 work · from 2016 · probable
Junnan LiArchitecture researcherCreated BLIP and BLIP-2, bootstrapping vision-language pretraining by bridging frozen image and text models.1 work · from 2023 · probable
Junyang LinArchitecture researcherCore contributor to the Qwen model family, building and training Alibaba's open large language and multimodal models.1 work · from 2024 · probable
Junyoung ChungArchitecture researcherWorked on AlphaCode, the system that generated and filtered candidate programs to reach competitive-programming performance.1 work · from 2022 · probable
Jyoti AnejaDataset creatorContributed to 'Textbooks Are All You Need', showing that curated textbook-quality training data lets small phi models punch above their size.1 work · from 2023 · probable
Kai ChenArchitecture researcherCo-authored 'Efficient Estimation of Word Representations', the word2vec work that made dense word embeddings cheap to train at scale.1 work · from 2013 · probable
Kai LiDataset creatorCo-authored ImageNet, the labeled image dataset whose scale enabled the deep-learning breakthrough in computer vision.1 work · from 2009 · probable
Kan WuTraining-method researcherWorked on FP8-LM / FP8 training, showing large models can be trained in 8-bit floating point to cut compute and memory.1 work · from 2023 · probable
Karan GoelArchitecture researcherCo-authored S4, the structured state-space sequence model that handles very long sequences as a subquadratic alternative to attention.1 work · from 2022 · probable
Karol HausmanSystems researcherWorked on SayCan ('Do As I Can, Not As I Say'), grounding language-model plans in a robot's real-world affordances for execution.1 work · from 2022 · probable
Kawin EthayarajhTraining-method researcherContributed to the preference-optimization family, formulating direct alignment objectives that fine-tune models from preference data without a separate reward model.1 work · from 2024 · probable
Kelvin GuuArchitecture researcherBuilt REALM, integrating a learned retriever into language-model pretraining to inject external knowledge.1 work · from 2020 · probable
Kevin ClarkArchitecture researcherCreated ELECTRA's replaced-token-detection pretraining, a more sample-efficient alternative to masked language modeling.1 work · from 2020 · probable
Kevin GimpelArchitecture researcherCo-authored ALBERT, which used parameter sharing and factorized embeddings to shrink BERT while retaining accuracy.1 work · from 2019 · probable
Kevin LinBenchmark creatorCo-authored 'Lost in the Middle', the study quantifying how long-context models degrade when relevant information sits mid-prompt.1 work · from 2023 · probable
Kevin MengTraining-method researcherDeveloped ROME, locating and directly editing factual associations stored in a model's weights.1 work · from 2022 · probable
Koray KavukcuogluTraining-method researcherCo-authored 'Human-level Control through Deep RL' (DQN), which combined Q-learning with deep networks to learn Atari from pixels.1 work · from 2015 · probable
Kristina ToutanovaArchitecture researcherCo-authored BERT, the masked-language-model pretraining that became the standard encoder for transfer learning in NLP.1 work · from 2018 · probable
Krzysztof ChoromanskiArchitecture researcherCo-authored Performers / linearized attention, approximating softmax attention with random features for linear-time sequence processing.1 work · from 2020 · probable
Krzysztof MaziarzArchitecture researcherCo-authored 'Outrageously Large Neural Networks', the sparsely-gated mixture-of-experts layer that scales capacity without proportional compute.1 work · from 2017 · probable
Kuldip PaliwalScientific originatorCo-originated bidirectional recurrent neural networks, letting sequence models condition on both past and future context.1 work · from 1997 · probable
Kurt HornikTheoretical researcherProved the universal approximation theorem, establishing that feedforward networks can approximate any continuous function.1 work · from 1989 · probable
Laurent SifreArchitecture researcherWorked on retrieval-augmented models FiD, RETRO and Atlas, coupling language models with external document retrieval for knowledge-grounded generation.1 work · from 2021 · probable
Leandro von WerraOpen-source implementerDrove open code-model efforts (BigCode / StarCoder), building the datasets and training pipelines for permissively-licensed code LLMs.1 work · from 2023 · probable
Lei ZhangArchitecture researcherWorked on open-vocabulary vision systems SAM, Grounding DINO and ImageBind for promptable segmentation and multimodal grounding.1 work · from 2023 · probable
Leopold AschenbrennerResearch communicatorWrote 'Situational Awareness,' an influential synthesis of AI scaling trajectories and their strategic implications.1 work · from 2024 · probable
Leslie GrovesExecutiveDirected the Manhattan Project, running the Los Alamos program that built the first atomic weapons.1 work · from 1945 · probable
Li DengArchitecture researcherCo-authored the deep neural network acoustic-modeling work that replaced Gaussian mixtures and brought deep learning into speech recognition.1 work · from 2012 · probable
Li-Jia LiDataset creatorCo-authored ImageNet, building the large labeled image collection that anchored the modern computer-vision benchmark.1 work · from 2009 · probable
Liang WangTraining-method researcherWorked on contrastive text embeddings (E5), training general-purpose text representations via large-scale contrastive learning.1 work · from 2019 · probable
Linxi FanArchitecture researcherCo-authored Voyager and generative-agent work, building LLM-driven agents that autonomously explore and acquire skills in open-ended environments.1 work · from 2023 · probable
Llion JonesArchitecture researcherCo-authored the Transformer architecture now standard across language and multimodal models.1 work · from 2017 · probable
Lucas BeyerArchitecture researcherCo-authored the Vision Transformer, applying a pure Transformer to image patches and displacing convolutions for large-scale vision.1 work · from 2021 · probable
Ludwig SchmidtDataset creatorCo-authored LAION-5B, the open billion-scale image-text dataset that made large multimodal training reproducible outside big labs.1 work · from 2022 · ✓ verified
Ludwig SchubertSafety researcherCo-authored 'Zoom In', the circuits interpretability work reverse-engineering individual neurons and features inside vision networks.1 work · from 2020 · probable
Lukasz KaiserArchitecture researcherCo-authored the Transformer, contributing to the attention design underlying current language models.1 work · from 2017 · ✓ verified
Łukasz KaiserArchitecture researcherWorked on linearized / efficient attention variants that reduce the Transformer's quadratic cost for long sequences.1 work · from 2020 · probable
Manaal FaruquiArchitecture researcherWorked on long-context attention mechanisms Ring Attention, Infini-Attention and MLA that extend context length under memory constraints.1 work · from 2024 · probable
Manzil ZaheerArchitecture researcherCo-authored Big Bird sparse attention, using a sparse attention pattern to process much longer sequences in linear complexity.1 work · from 2020 · probable
Marah AbdinOpen-source implementerContributed to Microsoft's open phi ecosystem, building the data curation and training behind small high-quality open models.1 work · from 2024 · probable
Margaret MitchellGovernance researcherCo-authored 'Stochastic Parrots' and developed model-documentation practices for responsible model release.1 work · from 2021 · probable
Marjan GhazvininejadArchitecture researcherCo-authored BART, the denoising sequence-to-sequence pretraining that unified generation and understanding tasks.1 work · from 2020 · probable
Matthew E. PetersArchitecture researcherWorked on sparse-attention methods that let Transformers scale to longer inputs without full quadratic attention.1 work · from 2020 · probable
Max WellingScientific originatorCo-authored 'Auto-Encoding Variational Bayes' (the VAE), establishing amortized variational inference as a foundation for deep generative models.1 work · from 2013 · probable
Mehdi MirzaArchitecture researcherCo-authored the original Generative Adversarial Networks paper and conditional GANs, seeding the adversarial generative-model line.1 work · from 2014 · probable
Melanie SubbiahArchitecture researcherCo-authored 'Language Models are Few-Shot Learners' (GPT-3), demonstrating in-context learning at scale.1 work · from 2020 · probable
Michael AhnArchitecture researcherCo-created SayCan, grounding language-model plans in a robot's real-world affordances.1 work · from 2022 · probable
Michael MatenaArchitecture researcherCo-authored T5 ('Exploring the Limits of Transfer Learning'), casting every NLP task as text-to-text under one pretrained model.1 work · from 2020 · probable
Michael PetrovSafety researcherCo-authored 'Zoom In', the interpretability work dissecting neurons and circuits inside vision models.1 work · from 2020 · probable
Michael PoliArchitecture researcherWorked on subquadratic sequence models RWKV, Hyena and attention-SSM hybrids as efficient alternatives to attention.1 work · from 2023 · probable
Michael S. BernsteinArchitecture researcherCo-authored generative-agents work, building LLM-driven simulated characters that plan, remember and interact in a sandbox world.1 work · from 2023 · probable
Michiel de JongArchitecture researcherCo-authored grouped-query attention (GQA), which trades attention heads for faster, memory-cheaper inference.1 work · from 2023 · probable
Mike SchusterArchitecture researcherCo-invented bidirectional recurrent neural networks, letting sequence models use both past and future context.1 work · from 1997 · probable
Mikhail PavlovArchitecture researcherCo-authored DALL-E ('Zero-Shot Text-to-Image Generation'), the autoregressive Transformer that generates images from text tokens.1 work · from 2021 · probable
Miles TurpinSafety researcherShowed chain-of-thought explanations can be unfaithful, misrepresenting the model's actual decision process.1 work · from 2023 · probable
Miljan MarticSafety researcherCo-authored 'Deep Reinforcement Learning from Human Preferences', the RLHF foundation for training agents from human comparison feedback.1 work · from 2017 · probable
Ming DingArchitecture researcherCo-authored GLM, an autoregressive blank-infilling pretraining objective underlying the GLM/ChatGLM model line.1 work · from 2021 · probable
Mingda ChenArchitecture researcherCo-authored ALBERT, using cross-layer parameter sharing to build a lighter BERT variant.1 work · from 2019 · probable
Mirac SuzgunBenchmark creatorWorked on BIG-bench, BBH and HELM, the broad evaluation suites that stress-test language-model capabilities across many tasks.1 work · from 2022 · probable
Mohammad BavarianTraining-method researcherWorked on fill-in-the-middle training for code models (CodeGen/InCoder line), enabling models to infill code given surrounding context.1 work · from 2022 · probable
Mohammad Gheshlaghi AzarTraining-method researcherDeveloped IPO, giving preference optimization a theoretical footing distinct from RLHF's reward modeling.1 work · from 2024 · probable
Mohammad NorouziArchitecture researcherCo-authored Imagen ('Photorealistic Text-to-Image Diffusion Models'), showing large text encoders paired with diffusion produce high-fidelity images.1 work · from 2022 · probable
Mohammad ShoeybiSystems researcherBuilt Megatron-LM, the tensor-parallel training system used to train many multi-billion-parameter language models.1 work · from 2019 · probable
Mostafa DehghaniTraining-method researcherCo-authored UL2, unifying denoising pretraining objectives so one model handles both understanding and generation.1 work · from 2023 · probable
Mostofa PatwaryInfrastructure builderCo-authored Megatron-LM, the tensor/model-parallel framework for training Transformers too large to fit on one GPU.1 work · from 2019 · probable
Mrinank SharmaSafety researcherWorked on reward-model overoptimization and sycophancy, characterizing failure modes where models game the reward or flatter users.1 work · from 2022 · probable
Myle OttTraining-method researcherCo-authored RoBERTa, showing that longer training on more data with tuned hyperparameters substantially improves BERT pretraining.1 work · from 2019 · probable
Nathaniel RochesterScientific originatorCo-organized the 1956 Dartmouth workshop that founded artificial intelligence as a field.1 work · from 1956 · probable
Neil RabinowitzTraining-method researcherCo-authored elastic weight consolidation ('Overcoming Catastrophic Forgetting'), protecting important weights so networks learn new tasks without erasing old ones.1 work · from 2017 · probable
Nelson F. LiuBenchmark creatorDemonstrated the lost-in-the-middle effect, where models underuse information placed mid-context.1 work · from 2023 · probable
Nicholas JosephSafety researcherCo-authored 'In-context Learning and Induction Heads', linking in-context learning to specific induction-head circuits in Transformers.1 work · from 2022 · probable
Nicholas MetropolisScientific originatorOriginated the Monte Carlo method and the Metropolis algorithm, the sampling foundation behind much of stochastic computation.1 work · from 1949 · probable
Nick CammarataSafety researcherCo-authored 'Zoom In', the circuits work interpreting individual features and their connections inside neural networks.1 work · from 2020 · probable
Nick RyderArchitecture researcherCo-authored 'Language Models are Few-Shot Learners' (GPT-3), the scaling result that demonstrated few-shot in-context learning.1 work · from 2020 · probable
Nicolas PapernotTheoretical researcherCo-authored 'The Curse of Recursion' (model collapse), showing that training successively on generated data degrades model quality.1 work · from 2024 · probable
Nikita KitaevArchitecture researcherCreated the Reformer, using locality-sensitive hashing to make attention memory-efficient.1 work · from 2020 · probable
Niklas MuennighoffDataset creatorStudied scaling under data constraints, quantifying returns from repeating limited training data.1 work · from 2023 · probable
Nils ReimersArchitecture researcherCreated Sentence-BERT, producing dense sentence embeddings suitable for fast semantic search.1 work · from 2019 · probable
Nisan StiennonTraining-method researcherCo-developed learning-to-summarize from human feedback, an early RLHF pipeline preceding InstructGPT.1 work · from 2020 · probable
Nitish SrivastavaTraining-method researcherCo-invented dropout, a regularization method that reduced overfitting across deep networks.1 work · from 2014 · probable
Noah BrownSystems researcherWorked on SayCan ('Do As I Can, Not As I Say'), connecting language-model plans to feasible robotic actions in the physical world.1 work · from 2022 · probable
Noah ShinnOpen-source implementerBuilt Reflexion, giving agents verbal self-reflection memory that improves performance across attempts.1 work · from 2023 · probable
Norbert WienerTheoretical researcherFounded cybernetics, framing feedback and control in machines and organisms that shaped early AI thinking.1 work · from 1948 · probable
Olatunji RuwaseInfrastructure builderCo-authored ZeRO, the DeepSpeed memory-partitioning optimizer that shards states across devices to train very large models.1 work · from 2020 · probable
Omar KhattabArchitecture researcherCreated ColBERT late-interaction retrieval, a widely-used efficient neural search method for RAG systems.1 work · from 2020 · ✓ verified
Omer LevyArchitecture researcherCo-authored BART, the denoising autoencoder pretraining that unified bidirectional encoding with autoregressive generation.1 work · from 2020 · probable
Oscar SainzBenchmark creatorDocumented pervasive benchmark data contamination undermining trust in LLM evaluation results.1 work · from 2023 · probable
Oskar MorgensternScientific originatorCo-authored 'Theory of Games and Economic Behavior', founding game theory and the utility framework behind rational-agent modeling.1 work · from 1944 · probable
Owain EvansBenchmark creatorWorked on contamination-resistant benchmarks, designing evaluations that resist leakage of test data into training sets.1 work · from 2021 · probable
Panupong PasupatArchitecture researcherCo-authored REALM, adding a learned retriever to language-model pretraining so the model can pull in external knowledge.1 work · from 2020 · probable
Paolo FrasconiTheoretical researcherCo-authored 'Learning Long-Term Dependencies is Difficult', analyzing why gradients vanish and RNNs struggle over long horizons.1 work · from 1994 · probable
Pascal VincentScientific originatorCo-authored 'A Neural Probabilistic Language Model', the early neural LM that learned distributed word representations jointly with prediction.1 work · from 2003 · probable
Patrice SimardTheoretical researcherCo-authored the 1994 'Learning Long-Term Dependencies is Difficult' analysis showing gradient-trained recurrent nets cannot capture long-range dependencies, the vanishing-gradient result that motivated the LSTM.1 work · from 1994 · probable
Patrick EsserArchitecture researcherCo-designed Latent Diffusion, moving the denoising process into a compressed latent space to make high-resolution text-to-image generation tractable.1 work · from 2022 · ✓ verified
Patrick HaffnerArchitecture researcherCo-authored LeNet, the gradient-trained convolutional network for document recognition that templated modern computer vision.1 work · from 1998 · probable
Patrick LewisScientific originatorIntroduced retrieval-augmented generation (RAG), the standard pattern for grounding language models in external documents.1 work · from 2020 · probable
Pauline LucArchitecture researcherCo-built Flamingo, the few-shot visual-language model that bridges frozen vision and language backbones with cross-attention.1 work · from 2022 · probable
Peilin ZhongArchitecture researcherCo-designed Titans, adding a learned neural long-term memory module to extend Transformer context beyond the attention window.1 work · from 2025 · probable
Peng ChengTraining-method researcherCo-developed FP8-LM, training large language models in 8-bit floating point to cut memory and compute cost.1 work · from 2023 · probable
Pengcheng HeArchitecture researcherDesigned DeBERTa's disentangled attention, which topped language-understanding benchmarks among encoder models.1 work · from 2021 · probable
Peter DayanTheoretical researcherCo-authored the 1992 proof of Q-learning convergence, a foundational result for value-based reinforcement learning.1 work · from 1992 · probable
Peter F. BrownScientific originatorPioneered statistical machine translation with the IBM alignment models, founding data-driven translation.1 work · from 1993 · probable
Peter J. LiuArchitecture researcherCo-authored T5, casting every NLP task as text-to-text and introducing the C4 web corpus used to train it.1 work · from 2020 · probable
Peter NorvigResearch communicatorCo-authored 'The Unreasonable Effectiveness of Data,' arguing large datasets beat elaborate models, a precursor to the scaling view.1 work · from 2009 · probable
Peter ShawArchitecture researcherIntroduced relative position representations for self-attention, an alternative to absolute positional encoding.1 work · from 2018 · probable
Pieter AbbeelSystems researcherCo-authored Ring Attention, distributing attention computation across devices to scale context length near-linearly with hardware.1 work · from 2024 · probable
Piotr DollárArchitecture researcherCo-created the Segment Anything Model (SAM), a promptable image-segmentation foundation model and its billion-mask dataset.1 work · from 2023 · probable
Piyush SharmaArchitecture researcherCo-designed ALBERT's factorized embeddings and cross-layer parameter sharing to shrink BERT while keeping accuracy.1 work · from 2019 · probable
Qingyang WuArchitecture researcherCo-built LLaVA, connecting a vision encoder to a language model through visual instruction tuning.1 work · from 2023 · probable
Qingyun WuSystems researcherCo-created AutoGen, a framework for orchestrating conversational multi-agent LLM systems with tool use and memory.1 work · from 2023 · probable
Quentin MalarticDataset creatorCo-produced the RefinedWeb corpus, showing filtered web text alone can train strong LLMs like Falcon.1 work · from 2023 · probable
Rafael RafailovTraining-method researcherLead author of Direct Preference Optimization, which aligns models from preferences without a separate reward model or RL loop.1 work · from 2023 · ✓ verified
Raia HadsellTraining-method researcherCo-authored Elastic Weight Consolidation for overcoming catastrophic forgetting, using Fisher-information penalties to protect prior tasks.1 work · from 2017 · probable
Raul PuriInfrastructure builderCo-built Megatron-LM's tensor-parallel training, splitting transformer layers across GPUs to train very large models.1 work · from 2019 · probable
Ray SolomonoffTheoretical researcherFounded algorithmic probability and inductive inference, a formal theory of prediction underlying modern views of learning.1 work · from 1964 · probable
Raymond LiDataset creatorHelped build StarCoder and its open code corpus, providing permissively sourced training data for code models.1 work · from 2023 · probable
Razvan PascanuTheoretical researcherCo-authored Elastic Weight Consolidation for continual learning without catastrophic forgetting of earlier tasks.1 work · from 2017 · probable
Reiichiro NakanoArchitecture researcherBuilt WebGPT, letting a model browse and cite web sources to answer long-form questions.1 work · from 2021 · probable
Réjean DucharmeScientific originatorCo-authored the 2003 Neural Probabilistic Language Model that introduced learned distributed word embeddings.1 work · from 2003 · probable
Rémi MunosTheoretical researcherReinforcement-learning theorist behind preference-optimization methods for aligning models from human preference data.1 work · from 2024 · probable
Rewon ChildArchitecture researcherCo-authored GPT-2 and Sparse Transformers, scaling decoder-only language models and sparse attention patterns.1 work · from 2019 · probable
Richard EvansArchitecture researcherCo-built AlphaFold2's Evoformer architecture, achieving atomic-accuracy protein structure prediction.1 work · from 2021 · probable
Richard VencuInfrastructure builderBuilt the data infrastructure behind LAION-5B, the open billion-scale image-text dataset used to train diffusion models.1 work · from 2022 · probable
Robert L. MercerScientific originatorCo-authored the IBM statistical machine-translation models that founded data-driven, alignment-based translation.1 work · from 1993 · probable
Roberto DessiTraining-method researcherCo-authored Toolformer, teaching a language model to self-supervise when and how to call external APIs.1 work · from 2023 · probable
Robin RombachArchitecture researcherCreated latent diffusion (Stable Diffusion), making high-quality open text-to-image generation efficient and widely accessible.1 work · from 2022 · ✓ verified
Rohan VarmaInfrastructure builderCo-built PyTorch Fully Sharded Data Parallel (FSDP), sharding parameters and optimizer state to train models beyond single-device memory.1 work · from 2023 · probable
Rohit GirdharArchitecture researcherCo-created ImageBind, learning a single joint embedding space that binds six modalities to images.1 work · from 2023 · probable
Romain BeaumontDataset creatorCo-built LAION-5B and its large-scale image-text retrieval and deduplication tooling.1 work · from 2022 · probable
Ronan CollobertArchitecture researcherShowed a single neural network could learn multiple NLP tasks from scratch, presaging unified representations.1 work · from 2011 · probable
Ross AndersonTheoretical researcherCo-authored 'The Curse of Recursion', analyzing how model collapse arises when models are trained recursively on generated data.1 work · from 2024 · probable
Ross GirshickArchitecture researcherCo-created the Segment Anything Model; earlier pioneered the R-CNN line that defined region-based object detection.1 work · from 2023 · ✓ verified
Ruibin XiongArchitecture researcherAnalyzed pre- versus post-layernorm placement, explaining and improving transformer training stability.1 work · from 2020 · probable
Rylan SchaefferTheoretical researcherArgued that many emergent abilities are artifacts of discontinuous metrics rather than genuine capability jumps.1 work · from 2023 · probable
Saining XieArchitecture researcherCo-designed the Diffusion Transformer (DiT), replacing the diffusion U-Net with a scalable transformer backbone.1 work · from 2022 · probable
Sam McCandlishTheoretical researcherCo-authored the scaling laws for language models that turned model sizing into a quantitative planning problem.1 work · from 2020 · probable
Sameer KumarSystems researcherCo-authored ST-MoE, stabilizing large sparse mixture-of-experts training and its expert-parallel execution at scale.1 work · from 2022 · probable
Samuel R. BowmanSafety researcherWorks on LLM evaluation and alignment, including studies of unfaithful chain-of-thought and reasoning correctives.1 work · from 2023 · probable
Samyam RajbhandariSystems researcherCreated ZeRO memory optimizations (DeepSpeed) that made trillion-parameter training possible on existing hardware.1 work · from 2020 · probable
Sandipan KunduSafety researcherCo-authored Constitutional AI, training a harmless assistant from a written set of principles using AI feedback.1 work · from 2022 · probable
Sang Michael XieDataset creatorDeveloped DoReMi, optimizing the mixture of pretraining data domains to improve downstream performance.1 work · from 2023 · probable
Sanmi KoyejoTheoretical researcherCo-authored 'Are Emergent Abilities of LLMs a Mirage?', arguing emergence is an artifact of discontinuous evaluation metrics.1 work · from 2023 · probable
Santiago FernándezTraining-method researcherCo-invented Connectionist Temporal Classification (CTC), the alignment-free loss enabling end-to-end sequence labeling for speech.1 work · from 2006 · probable
Saurabh SaxenaArchitecture researcherCo-built Imagen, showing large frozen text encoders drive photorealistic text-to-image diffusion.1 work · from 2022 · probable
Saurav KadavathSafety researcherCo-authored Constitutional AI and work on language models self-evaluating the reliability of their own answers.1 work · from 2022 · probable
Scott GraySystems researcherCo-authored DALL-E and built the block-sparse GPU kernels that made large sparse transformers practical.1 work · from 2021 · probable
Sebastian GoodmanArchitecture researcherCo-authored ALBERT, the factorized-embedding, parameter-sharing BERT variant that reduced model size.1 work · from 2019 · probable
Sébastien BubeckArchitecture researcherLed the phi small-model line and authored the widely debated Sparks of AGI analysis of GPT-4.1 work · from 2024 · probable
Sergey IoffeTraining-method researcherInvented batch normalization, which accelerated and stabilized training of deep networks across the field.1 work · from 2015 · probable
Sewon MinArchitecture researcherCo-authored Dense Passage Retrieval, learning dense dual-encoder embeddings for open-domain question answering.1 work · from 2020 · probable
Seymour PapertTheoretical researcherCo-authored the Perceptrons critique that mapped the limits of linear models and shaped the field's next decades.1 work · from 1969 · probable
Shan CarterSafety researcherCo-authored the 'Zoom In' Circuits work, visualizing individual features and their connections inside vision networks.1 work · from 2020 · probable
Shane LeggExecutiveCo-founded DeepMind and helped define AI safety as a research agenda alongside preference-learning work.1 work · from 2017 · probable
Shanghai AI LaboratoryInfrastructure builderOrganization producing open foundation models and evaluation toolchains (InternLM, OpenCompass) for the research ecosystem.1 work · from 2024 · probable
Shaohan HuangArchitecture researcherCo-authored open production vision-language and foundation models in the BEiT/Kosmos line.1 work · from 2023 · probable
Shengfeng PanArchitecture researcherCo-authored RoFormer, introducing rotary position embeddings (RoPE) now standard across large language models.1 work · from 2021 · probable
Shilong LiuArchitecture researcherCo-created Grounding DINO, an open-set object detector fusing language grounding with detection transformers.1 work · from 2023 · probable
Shitao XiaoArchitecture researcherCo-built the BGE contrastive text-embedding models widely used for dense retrieval.1 work · from 2019 · probable
Shouyuan ChenArchitecture researcherIntroduced position interpolation, cheaply extending pretrained RoPE context windows.1 work · from 2023 · probable
Shuyan ZhouBenchmark creatorCo-created WebArena, a realistic benchmark for evaluating LLM agents on multi-step web navigation tasks.1 work · from 2023 · probable
Sid BlackDataset creatorCo-produced The Pile open training corpus and the GPT-Neo open language models built on it.1 work · from 2021 · probable
Silvio SavareseArchitecture researcherCo-authored BLIP and BLIP-2, bootstrapping vision-language pretraining by connecting frozen image and language models.1 work · from 2023 · probable
Simon OsinderoScientific originatorCo-authored the 2006 deep belief network paper whose layer-wise pretraining reignited deep learning.1 work · from 2006 · probable
Siyuan ZhuangSystems researcherCo-built vLLM's PagedAttention, paging the KV cache like virtual memory to raise LLM serving throughput.1 work · from 2023 · probable
Song HanSystems researcherPioneered model compression and quantization (deep compression, AWQ) for efficient neural-network inference.1 work · from 2022 · probable
Stanislaw UlamScientific originatorCo-invented the Monte Carlo method, foundational to stochastic simulation and sampling.1 work · from 1949 · probable
Stephanie LinBenchmark creatorCreated TruthfulQA, a benchmark measuring whether models reproduce common human falsehoods.1 work · from 2021 · probable
Stephen A. Della PietraScientific originatorCo-authored the IBM statistical machine-translation alignment models that founded data-driven translation.1 work · from 1993 · probable
Steven BasartBenchmark creatorCo-created MMLU, the 57-subject multitask knowledge benchmark used to rank language-model capability.1 work · from 2021 · probable
Steven C. H. HoiArchitecture researcherCo-authored BLIP and BLIP-2, advancing vision-language pretraining with frozen-encoder bootstrapping.1 work · from 2023 · probable
Steven C.H. HoiArchitecture researcherCo-authored CodeT5 and related code language models, including fill-in-the-middle infilling for program synthesis.1 work · from 2022 · probable
Suchir BalajiTraining-method researcherCo-authored WebGPT, fine-tuning a browsing model with human feedback for long-form question answering.1 work · from 2021 · probable
Sumit ChopraArchitecture researcherCo-authored Memory Networks, augmenting neural networks with an addressable external memory for reasoning.1 work · from 2014 · probable
Sumit SanghaiArchitecture researcherCo-authored Grouped-Query Attention (GQA), sharing key/value heads to speed up inference-time attention.1 work · from 2023 · probable
Suriya GunasekarDataset creatorShowed high-quality textbook-like data lets small models achieve strong coding ability with less compute.1 work · from 2023 · probable
Tao LeiArchitecture researcherCo-authored expert-choice routing for mixture-of-experts, letting experts select tokens to balance load.1 work · from 2022 · probable
Tao XuArchitecture researcherCo-authored Whisper, training robust multilingual speech recognition on large-scale weakly-labeled audio.1 work · from 2022 · probable
Tatsunori B. HashimotoBenchmark creatorCo-created Chatbot Arena and LLM-as-judge evaluation, ranking chat models by crowd-sourced pairwise preference.1 work · from 2023 · probable
Teven Le ScaoArchitecture researcherContributor to BLOOM, the open multilingual large language model trained through the BigScience collaboration.1 work · from 2022 · probable
Thibaut LavrilArchitecture researcherCo-authored LLaMA, the efficient open foundation-model family that seeded much of open-weights research.1 work · from 2023 · probable
Thomas HubertTraining-method researcherCo-authored AlphaGo and AlphaZero, combining deep networks with Monte-Carlo tree search and self-play.1 work · from 2016 · probable
Thomas ScialomBenchmark creatorCo-authored agent and web benchmarks alongside tool-use work like Toolformer for evaluating LLM capabilities.1 work · from 2023 · probable
Thomas WolfInfrastructure builderCo-led BLOOM, the open multilingual LLM, and the Hugging Face Transformers ecosystem underpinning open-model research.1 work · from 2022 · probable
Tianle CaiSystems researcherCo-authored Medusa, adding parallel decoding heads to accelerate LLM inference via speculative decoding.1 work · from 2023 · probable
Tianqi ChenSystems researcherBuilt activation checkpointing for sublinear-memory training and core ML infrastructure widely reused across the ecosystem.1 work · from 2016 · probable
Tianyu GaoArchitecture researcherCo-authored SimCSE, a simple contrastive-learning method for producing strong sentence embeddings.1 work · from 2019 · probable
Tim BrooksArchitecture researcherCo-led Sora, framing video-generation models as learnable simulators of the physical world.1 work · from 2024 · probable
Tim GreenArchitecture researcherContributed to AlphaFold2's deep-learning system for highly accurate protein structure prediction.1 work · from 2021 · probable
Tim SalimansTraining-method researcherCo-authored GPT-1's generative pre-training approach to language understanding.1 work · from 2018 · probable
Timnit GebruGovernance researcherCo-authored 'Stochastic Parrots' and pushed the field to confront bias and accountability in large language models.1 work · from 2021 · probable
Timo SchickScientific originatorBuilt Toolformer, showing language models can teach themselves when and how to call external tools.1 work · from 2023 · probable
Timothée LacroixArchitecture researcherCo-built the Mistral 7B and Mixtral open-weight models emphasizing efficient attention and sparse mixture-of-experts.1 work · from 2023 · probable
Tom GoldsteinBenchmark creatorWorked on contamination-resistant benchmarks that measure model performance without training-set leakage.1 work · from 2021 · probable
Tomas MikolovScientific originatorBuilt Word2Vec, which made dense word embeddings the standard input representation for NLP.1 work · from 2013 · ✓ verified
Tommy FlowersSystems researcherEngineered Colossus, the first programmable electronic computer, for Bletchley Park codebreaking.1 work · from 1943 · probable
Tony LeeBenchmark creatorContributed to the BIG-bench, BBH, and HELM evaluation suites for measuring language-model capabilities.1 work · from 2022 · probable
Toran Bruce RichardsOpen-source implementerCreated AutoGPT, an early autonomous agent framework that popularized self-directed LLM loops.1 work · from 2023 · probable
Trenton BrickenSafety researcherCo-authored monosemanticity work using sparse autoencoders to extract interpretable features from models.1 work · from 2023 · probable
Tristan HumeSafety researcherCo-authored Toy Models of Superposition, characterizing how neural networks pack features into overlapping directions for interpretability.1 work · from 2022 · probable
Tsendsuren MunkhdalaiArchitecture researcherWorked on long-context attention mechanisms including Infini-Attention for unbounded-length sequence processing.1 work · from 2024 · probable
Urvashi KhandelwalArchitecture researcherDeveloped nearest-neighbor and late-interaction retrieval methods (kNN-LM, ColBERT-style rerankers) for augmenting language models.1 work · from 2020 · probable
Vahab MirrokniTheoretical researcherWorked on the Titans architecture introducing learned memory modules for long-context sequence models.1 work · from 2025 · probable
Vannevar BushResearch communicatorDescribed the Memex vision of associative information machines and built the US framework for funding large-scale science.1 work · from 1945 · probable
Ves StoyanovArchitecture researcherCo-authored BART's denoising sequence-to-sequence pre-training objective.1 work · from 2020 · probable
Vinay RamaseshTraining-method researcherWorked on Minerva, adapting large language models to solve quantitative reasoning problems.1 work · from 2022 · probable
Vincent J. Della PietraScientific originatorCo-authored the IBM statistical machine translation models that formalized MT as a probabilistic estimation problem.1 work · from 1993 · probable
Vineet KosarajuBenchmark creatorContributed to the GSM8K and MATH datasets for evaluating mathematical reasoning.1 work · from 2021 · probable
Vinh Q. TranTraining-method researcherWorked on UL2's unified pre-training objective mixing denoising modes across model types.1 work · from 2023 · probable
Vinod NairArchitecture researcherCo-introduced rectified linear units, the activation that made deep networks far easier to train.1 work · from 2010 · probable
Vladimir KarpukhinArchitecture researcherCreated Dense Passage Retrieval, replacing sparse lexical search with learned dense embeddings for open-domain QA.1 work · from 2020 · probable
Volodymyr MnihArchitecture researcherBuilt the Deep Q-Network that first learned Atari games from pixels, launching deep reinforcement learning.1 work · from 2015 · probable
von NeumannScientific originatorDescribed the stored-program computer in the EDVAC report, defining the architecture nearly all general-purpose digital computers still follow.1 work · from 1945 · probable
Walter PittsTheoretical researcherCo-formulated the McCulloch-Pitts neuron, showing networks of simple units could compute logical functions.1 work · from 1943 · probable
Warren McCullochScientific originatorCo-authored the 1943 model of the artificial neuron as a logic unit, the starting point for neural computation.1 work · from 1943 · probable
Wei DongDataset creatorContributed to building the ImageNet dataset that anchored the deep-learning vision era.1 work · from 2009 · probable
Wei LiTraining-method researcherCo-authored T5, framing all NLP tasks as text-to-text transfer learning.1 work · from 2020 · probable
Wei-Lin ChiangBenchmark creatorBuilt Chatbot Arena and the LLM-as-Judge methodology for human-preference model evaluation.1 work · from 2023 · probable
Wenhu ChenTraining-method researcherDeveloped structured-reasoning methods such as program-of-thought for decomposing multi-step problems.1 work · from 2023 · probable
William ChanArchitecture researcherCo-authored Imagen, the cascaded diffusion model for photorealistic text-to-image generation.1 work · from 2022 · probable
William El SayedArchitecture researcherCo-founded Mistral and contributed to the Mistral 7B and Mixtral open-weight models.1 work · from 2023 · probable
William PeeblesArchitecture researcherDesigned the Diffusion Transformer (DiT), replacing the U-Net backbone in latent diffusion with a transformer.1 work · from 2022 · probable
William W. CohenTheoretical researcherWorked on structured-reasoning approaches integrating symbolic knowledge with neural language models.1 work · from 2023 · probable
Wojciech ZarembaBenchmark creatorCo-authored the HumanEval and MBPP benchmarks for evaluating code-generation models.1 work · from 2021 · probable
Woosuk KwonSystems researcherBuilt PagedAttention and vLLM, the memory-efficient serving system now widely used to deploy large language models.1 work · from 2023 · ✓ verified
Xavier GarciaTraining-method researcherWorked on UL2's mixture-of-denoisers pre-training objective.1 work · from 2023 · probable
Xi ChenArchitecture researcherContributed to open production vision-language models integrating image encoders with language backbones.1 work · from 2023 · probable
Xiaodong LiuArchitecture researcherCo-authored DeBERTa's disentangled attention and enhanced mask decoding.1 work · from 2021 · probable
Xingyao WangSystems researcherBuilt the SWE-agent and OpenHands frameworks for autonomous software-engineering agents.1 work · from 2024 · probable
Xinyun ChenTraining-method researcherWorked on program synthesis with execution feedback and unit-test-guided code generation.1 work · from 2022 · probable
Xu JiangTraining-method researcherContributed to InstructGPT, aligning language models to instructions via reinforcement learning from human feedback.1 work · from 2022 · probable
Xuanyi DongDataset creatorCo-authored DoReMi, optimizing pre-training data mixtures via domain reweighting.1 work · from 2023 · probable
Xuechen LiBenchmark creatorContributed to Chatbot Arena and LLM-as-Judge automatic evaluation of instruction-following models.1 work · from 2023 · probable
Yaniv LeviathanSystems researcherIntroduced speculative decoding, using a small draft model to accelerate large-model inference without changing the output distribution.1 work · from 2023 · probable
Yanli ZhaoSystems researcherLed PyTorch FSDP, giving practitioners memory-efficient fully-sharded data-parallel training for models too large to fit on one device.1 work · from 2023 · probable
Yann DuboisBenchmark creatorWorked on Chatbot Arena and LLM-as-Judge preference-based evaluation of language models.1 work · from 2023 · probable
Yann LeCunScientific originatorDesigned LeNet, the convolutional network that established gradient-trained CNNs as the template for modern computer vision.1 work · from 1998 · ✓ verified
Yarin GalTheoretical researcherCo-authored The Curse of Recursion, analyzing model collapse when models train on generated data.1 work · from 2024 · probable
Ye XiaArchitecture researcherCo-authored ALIGN, scaling contrastive image-text pre-training on noisy web-scale data.1 work · from 2021 · probable
Yee-Whye TehScientific originatorCo-authored deep belief nets and the dimensionality-reduction work that helped launch deep learning.1 work · from 2006 · probable
Yelong ShenTraining-method researcherCo-authored LoRA, the low-rank adaptation method for parameter-efficient fine-tuning.1 work · from 2022 · probable
Yi ZhangDataset creatorCo-authored Textbooks Are All You Need, showing high-quality curated data yields strong small models.1 work · from 2023 · probable
Yinfei YangArchitecture researcherCo-authored ALIGN's large-scale noisy image-text contrastive pre-training.1 work · from 2021 · probable
Yiren ZhaoTheoretical researcherCo-authored model-collapse analysis of degradation from training on synthetic data.1 work · from 2024 · probable
Yizhong WangTraining-method researcherWorked on instruction-tuning and self-generated instruction data for zero-shot task generalization.1 work · from 2022 · probable
Yoav LevineArchitecture researcherWorked on tool-use and function-calling paradigms (MRKL) connecting language models to external systems.1 work · from 2023 · probable
Yohei NakajimaOpen-source implementerCreated BabyAGI, an early autonomous task-driven agent loop in the AutoGPT lineage.1 work · from 2023 · probable
Yonatan BelinkovSafety researcherWorked on knowledge editing and interpretability of where facts are stored in language models.1 work · from 2022 · probable
Yong Jae LeeArchitecture researcherCo-authored LLaVA, connecting a vision encoder to a language model via visual instruction tuning.1 work · from 2023 · probable
Yossi MatiasSystems researcherWorked on speculative decoding methods (Medusa, EAGLE) for accelerating language-model inference.1 work · from 2023 · probable
Yu LuArchitecture researcherCo-authored RoFormer, introducing rotary position embeddings for the transformer.1 work · from 2021 · probable
Yu MengTraining-method researcherWorked on the preference-optimization family of alignment objectives beyond RLHF.1 work · from 2024 · probable
Yuan CaoTraining-method researcherCo-authored ReAct, interleaving reasoning traces with actions for tool-using agents.1 work · from 2023 · probable
Yuandong TianArchitecture researcherWorked on context-extension methods for scaling transformer context windows.1 work · from 2023 · probable
Yue WangArchitecture researcherBuilt CodeT5 and related code models, enabling pretrained encoder-decoder systems for program synthesis and code understanding.1 work · from 2022 · probable
Yuhui LiSystems researcherAuthored EAGLE, a speculative-decoding method for faster language-model inference.1 work · from 2023 · probable
Yujia LiArchitecture researcherLed AlphaCode, showing that large-scale sampling and filtering could reach competitive-programming-level code generation.1 work · from 2022 · probable
Yujie QianArchitecture researcherContributed to GLM, the general language model with autoregressive blank infilling.1 work · from 2021 · probable
Yunchang YangTheoretical researcherCo-authored the analysis of layer normalization placement (pre-LN) that stabilized transformer training.1 work · from 2020 · probable
Yunfeng LiuArchitecture researcherCo-authored RoFormer's rotary position embedding for the transformer.1 work · from 2021 · probable
Yuntao BaiSafety researcherLead author of Constitutional AI, replacing much human feedback with AI feedback against written principles.1 work · from 2022 · probable
Yury ZemlyanskiyArchitecture researcherWorked on grouped-query attention (GQA) for memory-efficient transformer inference.1 work · from 2023 · probable
Yuxiong HeInfrastructure builderCo-authored ZeRO, the DeepSpeed memory-partitioning technique for training very large models.1 work · from 2020 · probable
Zakhar ShumaylovTheoretical researcherCo-authored The Curse of Recursion, analyzing model collapse from recursive training on generated data.1 work · from 2024 · probable
Zhe ChenArchitecture researcherContributed to open production vision-language models (InternVL-style) scaling vision encoders.1 work · from 2023 · probable
Zheng LiuArchitecture researcherWorked on contrastive text-embedding models for dense retrieval.1 work · from 2019 · probable
Zhengxiao DuArchitecture researcherCo-created the GLM autoregressive blank-infilling pretraining objective behind the open GLM model family.1 work · from 2021 · probable
Zhenzhong LanArchitecture researcherLed ALBERT, using parameter sharing and factorized embeddings to make BERT-style pretraining far more parameter-efficient.1 work · from 2019 · probable
Zhifeng ChenInfrastructure builderCo-authored Mesh-TensorFlow and GSPMD for distributing large-model training across device meshes.1 work · from 2018 · probable
Zhihong ShaoTraining-method researcherContributed to DeepSeek-R1, eliciting reasoning in language models via reinforcement learning.1 work · from 2025 · probable
Zhuohan LiSystems researcherCo-authored PagedAttention and vLLM, the KV-cache paging system for high-throughput LLM serving.1 work · from 2023 · ✓ verified
Zihang DaiArchitecture researcherWorked on relative position representations in self-attention for the transformer.1 work · from 2018 · probable
Zora TungArchitecture researcherContributed to REALM, adding a learned retriever to language-model pre-training.1 work · from 2020 · probable
Andrej KarpathyOpen-source implementernanoGPT, char-rnn, deep learning education0 works · ✓ verified
Christopher OlahOpen-source implementerMechanistic interpretability0 works · ✓ verified
Farnoosh HashemiOpen-source implementerMemory consolidation in language models0 works · ✓ verified
Lilian WengOpen-source implementerRLHF, deep learning blog, agents0 works · ✓ verified
Phil WangOpen-source implementerOpen-source reimplementations of Transformers, DALL-E, etc.0 works · ✓ verified
Ross WightmanOpen-source implementertimm (PyTorch Image Models)0 works · ✓ verified
Sam AltmanOpen-source implementerOpenAI leadership; “Moore’s Law for Everything”0 works · ✓ verified
Soumith ChintalaOpen-source implementerPyTorch, DCGAN0 works · ✓ verified
Xinyang GengOpen-source implementerKoala, EasyLM, OpenLLaMA0 works · ✓ verified
Yangqing JiaOpen-source implementerCaffe deep learning framework0 works · ✓ verified