Reference Series · People Behind the Models 🧑‍🔬

The AI Researcher Atlas: 50 People Who Built the Field

Every lineage article on this site names a model. This one names the people. A tiered survey of roughly fifty researchers whose work underlies the labs this project has covered — foundational figures whose theories predate the current boom, the architects behind specific breakthroughs like the Transformer and RLHF, and a wave of specialists in interpretability, safety, and robotics defining the field's next decade. 2025-2026 has been an unusually volatile year for where these people actually work — several changed employers multiple times since our last research pass — so every current-role claim here is dated and confidence-flagged rather than stated as a permanent fact.

FrontierAGI Team

September 19, 2026 · 65 min read

Opening How to Read This Atlas Tier 1: Foundational The Transformer's 8 Authors Scaling Laws & GPT Safety & Alignment Other Architects Tier 3: Rising & Specialized Addendum Patterns Across the 50 Cross-Reference Index What's Missing References Videos Verdict

The Lineages Have Authors

Every GPT, Claude, Gemini, Grok, DeepSeek, Llama, and Mistral release this project has documented was built by identifiable people, working inside identifiable institutional lineages — advisor trees, prior labs, prior papers. This atlas surfaces those people directly, organized by contribution rather than current employer, because in 2025-2026 employer has become the least stable fact about many of them.

How to Read This Atlas

Selection Method and Confidence Tags

Selection Is Contribution-Weighted, Not a Popularity Ranking This list was built around documented technical contributions and clear influence on the labs and models this site already covers — not social-media following, citation count alone, or media prominence. It skews toward people whose work is directly traceable to specific breakthroughs in this project's other articles. It is not exhaustive, and reasonable people would swap several names in or out; the goal is a representative map, not a definitive ranking.

Every entry below carries an implicit confidence level. Where research turned up conflicting or single-sourced claims about a person's most recent move, this article says so explicitly using inline flags — Confirmed for claims cross-verified across multiple independent, reliable sources, and Unverified for claims that rest on a single aggregator source or conflicting reports, which should be checked directly before being repeated elsewhere.

Tier 1: Foundational

The Figures Whose Theories Predate the Boom

Geoffrey Hintonb. 1947, UK/Canada

Path: PhD Edinburgh (1978) → Carnegie Mellon → University of Toronto → Google (2013–2023, via DNNresearch acquisition).
Key Achievements: Co-invented/popularized backpropagation (1986, with Rumelhart and Williams); co-invented Boltzmann Machines (1983, with Sejnowski); supervised AlexNet (2012, with Sutskever and Krizhevsky), the CNN that won ImageNet and triggered the deep learning boom; 2018 Turing Award with Bengio and LeCun.
Current Focus: Left Google specifically to speak freely about AI risk; now an independent AI-safety commentator warning about job displacement and autonomous-agent risk. Confirmed

Yoshua Bengiob. 1964, France/Canada

Path: PhD McGill (1991) → founded Mila (1993) → Université de Montréal professor → founded LawZero (2025).
Key Achievements: Founded Mila, now one of the largest academic deep-learning hubs; co-developed early neural probabilistic language models underpinning modern NLP; 2018 Turing Award; advised Ian Goodfellow's GAN work.
Current Focus: Founded LawZero (June 2025), a nonprofit pursuing non-agentic "Scientist AI"; chaired the International AI Safety Report 2026, a 100+-expert assessment concluding AI capability progress still outpaces safety mitigation. Confirmed

Yann LeCunb. 1960, France

Path: PhD Université Pierre et Marie Curie (1987) → Bell Labs → NYU (2003–present) → Meta Chief AI Scientist (2013–2025).
Key Achievements: Developed convolutional neural networks and the LeNet architecture (late 1980s); 2019 Turing Award; built and led Meta's FAIR lab; championed "world models" and JEPA as an alternative to pure LLM scaling.
Current Focus: Announced departure from Meta in November 2025 (covered in this site's Meta lineage article) to found a new "world models" research venture, reportedly possibly Paris-based. No confirmed company name, funding, or launch date exists yet. Unverified (venture specifics)

Ilya SutskeverRussia/Israel/Canada

Path: PhD University of Toronto under Geoffrey Hinton → Google Brain → OpenAI co-founder/Chief Scientist (2015–2024) → Safe Superintelligence Inc. (SSI), 2024–present.
Key Achievements: Co-developed AlexNet (2012) with Krizhevsky under Hinton; co-authored the foundational sequence-to-sequence learning paper (2014); central to OpenAI's GPT research direction as co-founder and Chief Scientist.
Current Focus: CEO of SSI, which received a reported $5 billion investment commitment from Nvidia plus Vera Rubin platform access (announced July 2026), at a reported ~$32B valuation with no shipped product yet. Confirmed (Nvidia deal) Unverified (specific product timeline)

Andrej KarpathySlovakia/Canada

Path: PhD Stanford → OpenAI founding team (2015) → Tesla Senior Director of AI (2017–2022) → OpenAI (2023) → founded Eureka Labs (2024).
Key Achievements: Built and led Tesla's Autopilot/FSD computer-vision stack; highly influential AI educator (Stanford CS231n, "nanoGPT," "llm.c"); co-founded OpenAI's early research culture.
Current Focus: Founded Eureka Labs, an "AI-native school." A single low-tier aggregator source claims he joined Anthropic in May 2026 to lead a pretraining team — this could not be corroborated against any primary source and should be treated as unverified, possibly inaccurate. Unverified

Demis Hassabisb. 1976, UK

Path: Child chess prodigy/game designer → PhD Cognitive Neuroscience, UCL (2009) → co-founded DeepMind (2010) → CEO, Google DeepMind (post-2023 merger).
Key Achievements: Co-founded DeepMind with Shane Legg and Mustafa Suleyman; led AlphaGo (2016); led AlphaFold, solving the ~50-year protein-structure-prediction challenge; founded Isomorphic Labs.
Current Focus: CEO, Google DeepMind, continuing to oversee the unified Google AI research organization detailed in this site's Gemini lineage article. Awarded the 2024 Nobel Prize in Chemistry for AlphaFold. Confirmed

Shane LeggNew Zealand

Path: Academic AI/AGI-theory research → co-founded DeepMind (2010) → Chief AGI Scientist, Google DeepMind.
Key Achievements: Co-founded DeepMind; earlier academic work on formal/universal measures of intelligence; lead author of "Levels of AGI: Operationalizing Progress on the Path to AGI" (arXiv:2311.02462), the six-level framework covered in this site's AGI Benchmarks article.
Current Focus: Chief AGI Scientist, Google DeepMind; publicly discusses AGI timeline estimates in interviews. Specific timeline figures attributed to him should be checked against primary transcripts before repeating. Confirmed (role)

Dario AmodeiUSA

Path: Biophysics PhD (Princeton/Stanford) → Google Brain → OpenAI VP of Research (led GPT-2/GPT-3 efforts) → co-founded Anthropic (2021), CEO.
Key Achievements: Led research producing GPT-2 and GPT-3 at OpenAI; contributed to scaling-laws research; co-founded Anthropic, positioning it around Constitutional AI and interpretability, detailed fully in this site's Claude lineage article.
Current Focus: CEO of Anthropic, reportedly valued around $965 billion as of a May 2026 round (verify independently before citing); publicly pushing for independent AI oversight per Sept 2026 commentary. Unverified (exact valuation)

Daniela Amodeib. 1987, USA

Path: UC Santa Cruz (literature) → Capitol Hill staffer → Stripe (founding recruiter) → OpenAI VP of Safety & Policy, then VP of Research → co-founded Anthropic (2021), President.
Key Achievements: Held VP of Safety and Policy at OpenAI overseeing technical safety/policy implementation; co-founded Anthropic with brother Dario; leads research, engineering, product, governance and commercial execution as President.
Current Focus: President of Anthropic; publicly emphasizes the value of humanities training in AI hiring. Confirmed

Arthur Menschb. 1992, France

Path: École Polytechnique, ENS, PhD Université Paris-Saclay → DeepMind Paris (2020–2023) → co-founded Mistral AI (2023), CEO.
Key Achievements: Co-authored the Chinchilla compute-optimal scaling-laws paper at DeepMind; contributed to the Flamingo vision-language model; co-founded Mistral AI, detailed fully in this site's Mistral lineage article.
Current Focus: CEO of Mistral AI, valued at roughly €21B after its September 2026 Samsung-led round (see this site's Mistral lineage article for full funding history). Confirmed

Fei-Fei LiChina/USA

Path: PhD Caltech → Stanford faculty → Google Cloud AI Chief Scientist (2017-18) → co-founded Stanford HAI (2019) → founded World Labs (2024).
Key Achievements: Created ImageNet (2009), the labeled-image dataset that catalyzed the computer-vision/deep-learning revolution; co-founded Stanford HAI, whose AI Index this site's companion article covers in detail; longtime advocate for human-centered AI development.
Current Focus: Founder of World Labs, pursuing "spatial intelligence" and 3D world models; the company raised roughly $1 billion in a February 2026 round (Autodesk, AMD, Nvidia among investors), reaching a reported ~$5B valuation, and shipped its first product, Marble. Confirmed

Andrew NgUK/Hong Kong/Singapore/USA

Path: PhD UC Berkeley → Stanford faculty → co-founded Google Brain (2011) → co-founded Coursera (2012) → Chief Scientist, Baidu (2014-2017) → founded deeplearning.ai, Landing AI, AI Fund.
Key Achievements: Co-founded Google Brain, a foundational large-scale deep-learning research team; co-founded Coursera, helping popularize mass AI education; led Baidu's AI Group expanding deep learning at scale in China.
Current Focus: Spread across Founder of DeepLearning.AI, Managing General Partner of AI Fund, Executive Chairman of Landing AI, and Chairman/co-founder of Coursera, alongside an adjunct Stanford role. Confirmed

The Transformer's 8 Authors

One 2017 Paper, Eight Very Different 2026 Careers

Every model in every lineage article on this site descends from a single 2017 paper, "Attention Is All You Need." Its eight authors are worth tracking as a group precisely because their post-Google career paths diverged so dramatically — from returning to the exact frontier-lab race their paper enabled, to abandoning AI research for biotech, to founding a blockchain protocol.

Ashish VaswaniLead author

Led the Transformer's design at Google Brain. Co-founded Adept AI (2022), then Essential AI (2023) as CEO; Essential AI was reportedly acqui-hired by Nvidia in June 2026. Unverified (Nvidia role specifics)

Noam ShazeerScaling architect

Invented Mixture-of-Experts and multi-query attention techniques underlying the paper. Founded Character.AI (2021), was rehired by Google as Gemini co-technical-lead (2024) after Google's $2.7B deal for Character.AI's technology, then reportedly left for OpenAI in June 2026 to lead architecture research. Confirmed through mid-2026

Niki ParmarEncoder-decoder design

Co-founded Adept AI and Essential AI alongside Vaswani. Reportedly now at Anthropic, though the exact role and date are only lightly sourced. Unverified

Jakob UszkoreitAttention formulation

Left AI/NLP research entirely to found Inceptive (2021), applying generative deep learning to mRNA and drug design — the most complete field-change of any Transformer co-author.

Llion JonesCredited with the paper's title

Co-founded Sakana AI (Tokyo, 2023), pursuing nature-inspired/evolutionary approaches to model-building such as model merging, an explicit alternative research bet to pure scaling.

Aidan GomezGoogle Brain intern at 20

Co-founded and remains CEO of Cohere, which reportedly acquired Aleph Alpha in April 2026, combining into an entity valued near $20B; publicly critical of proposed cross-lab "AI safety cartel" coordination in September 2026 commentary. Confirmed

Łukasz KaiserTensor2Tensor library

Joined OpenAI in 2021 and reportedly continues contributing to reasoning-model research there as of 2026, though the specific current project is not independently confirmed.

Illia PolosukhinCo-author

Pivoted entirely to blockchain, co-founding NEAR Protocol (2018). Now a vocal advocate for decentralized/user-owned AI, though a specific 2026 regulatory claim he has cited in support of that thesis could not be independently corroborated here. Unverified (specific regulatory claim)

Scaling Laws & GPT

The People Who Made Scale the Strategy

Jared KaplanScaling laws, Anthropic co-founder

Physics PhD (Harvard); lead author of "Scaling Laws for Neural Language Models" (2020), the power-law relationships between model size, data, and compute that underpin the entire industry's scaling strategy — the same laws referenced throughout this site's GPT and lineage articles. Co-founded Anthropic as Chief Science Officer, helped develop Constitutional AI. Confirmed

Alec RadfordGPT, CLIP, Whisper lead author

Lead author of the original GPT (2018) and GPT-2 (2019) papers establishing generative pretraining as the dominant LLM paradigm; co-created CLIP, underlying most modern multimodal systems; led Whisper. Left OpenAI in December 2024; reportedly an advisor to Mira Murati's Thinking Machines Lab. Unverified (current advisory status)

John SchulmanInvented PPO

PhD Berkeley under Pieter Abbeel; invented Proximal Policy Optimization (2017), the RL algorithm central to RLHF and ChatGPT's alignment training. Co-founded OpenAI, moved to Anthropic in August 2024, then left Anthropic after just five months (February 2025) to become Chief Scientist at Thinking Machines Lab. Confirmed — note this corrects a common simplified "moved to Anthropic" summary; the full path involves a third stop.

Mira MuratiFormer OpenAI CTO

Oversaw shipping of ChatGPT, GPT-4, DALL-E, Codex, and Whisper as OpenAI CTO; briefly served as interim CEO during the November 2023 board crisis. Left OpenAI in September 2024 to found Thinking Machines Lab, which raised $2B at a $12B valuation in July 2025. Reports on a further raise conflict: one round of talks (targeting $50B valuation) reportedly collapsed by January 2026, while later reporting describes renewed talks near a $40B+ valuation with Nvidia funding roughly half, as of September 2026. Unverified (current funding status — actively in motion)

Safety & Alignment

The People Who Built the Guardrails

François CholletKeras creator, ARC-AGI

Created Keras (2015), the most widely-adopted high-level deep learning API. Created the ARC-AGI benchmark (2019), covered in depth in this site's AGI Benchmarks article, reframing AGI evaluation around sample-efficient abstraction rather than raw task performance. Left Google in November 2024 and co-founded Ndea (announced January 2025) with Zapier co-founder Mike Knoop, betting on program-synthesis approaches to AGI; went through Y Combinator's Winter 2026 batch. Confirmed

Jan LeikeSuperalignment co-lead

Co-led OpenAI's Superalignment initiative with Ilya Sutskever, targeting a 4-year alignment-solving timeline. His May 2024 public resignation letter, citing safety-culture concerns about OpenAI prioritizing shipping over safety, became a widely-cited flashpoint. Joined Anthropic (May 2024) as VP of Alignment Science, a role he continues to hold as of the most recent available reporting. Confirmed

Paul ChristianoCo-developed RLHF

Co-developed Reinforcement Learning from Human Feedback at OpenAI (2017), the technique underlying ChatGPT's alignment and now industry-standard post-training. Founded the Alignment Research Center (2021). Appointed Head of AI Safety at the US AI Safety Institute (April 2024), continuing in its renamed successor (CAISI). In September 2026, reportedly also appointed to OpenAI's own Foundation Board Safety and Security Committee while remaining at CAISI — a notable governance crossover between a government evaluator and the lab it evaluates, worth flagging as unusual and recent enough to verify directly. Unverified (dual-role specifics)

Other Architects

Foundational Techniques and 2026's Talent Churn

Ian GoodfellowInvented GANs

PhD Université de Montréal under Yoshua Bengio; invented Generative Adversarial Networks (2014), one of the most consequential generative-modeling ideas before diffusion models. Co-authored the field's standard "Deep Learning" textbook. Career spanned Google Brain, OpenAI, Apple, and Google DeepMind. A single source reports he co-founded a venture fund ("224 Ventures") in April 2026 alongside Yann LeCun and Oriol Vinyals — this specific claim needs a second corroborating source before being treated as fact. Unverified

Percy LiangFoundation models, HELM

PhD UC Berkeley; Stanford professor. Founded Stanford's Center for Research on Foundation Models and helped coin the term "foundation models" (2021). Created HELM, a widely-used standardized LLM evaluation suite. Continues as Founding Director of CRFM as of 2026, with no reported departure or major role change — one of the most stable profiles in this entire atlas. Confirmed

Timnit GebruAI ethics, DAIR founder

PhD Stanford under Fei-Fei Li. Co-authored "On the Dangers of Stochastic Parrots" (2021), a landmark critique of scaling LLMs without addressing bias and documentation. Departed Google in December 2020 amid a dispute over that paper's publication. Founded and continues to lead the Distributed AI Research Institute (DAIR); published a new book on the AI industry in 2026. Confirmed

Noam BrownSuperhuman poker AI, test-time compute

PhD Carnegie Mellon; built Libratus and Pluribus, superhuman poker-playing AIs using counterfactual regret minimization. Co-led Cicero at Meta, the first AI to reach human-level performance in the negotiation game Diplomacy. At OpenAI, a foundational contributor to the o1/o3 reasoning models' test-time compute scaling approach. A claim that his team solved a Millennium Prize Problem using a 10,000-agent system in 2026 is an extraordinary, single-sourced assertion this article does not treat as confirmed. Unverified (Millennium Prize claim — treat with significant skepticism)

Barret ZophNeural Architecture Search

Pioneered Neural Architecture Search at Google Brain. Reported 2024-2026 path is unusually turbulent: OpenAI VP of Research → co-founded Thinking Machines Lab as CTO with Murati (Oct 2024) → reportedly exited amid contested allegations, rejoined OpenAI (Jan 2026) → departed OpenAI again (~June 2026) → reportedly joined Google as VP of Research (~Aug 2026). Given four reported employer changes in under two years, each leg of this path should be independently verified before being cited as settled fact. Unverified (multiple legs)

Oriol VinyalsAlphaStar, Gemini co-lead

PhD UC Berkeley; led AlphaStar (2019), the first AI to reach Grandmaster level in StarCraft II. Rose to VP of Research and Gemini co-technical-lead at Google DeepMind. A cluster of reports describes him departing Google in August 2026 alongside Jeff Dean, Sanjay Ghemawat, and Quoc Le to found a new company — a headline-grade claim about a mass senior departure that this article treats as reported but not yet confirmed by an official primary source. Unverified — significant claim, seek primary confirmation

Cross-Reference: DeepSeek and Mistral Founders Liang Wenfeng (DeepSeek's founder) and Guillaume Lample and Timothée Lacroix (Mistral's co-founders alongside Arthur Mensch) are covered in full in this site's dedicated DeepSeek and Mistral lineage articles rather than repeated here — see the reference links below.

Tier 3: Rising & Specialized

Interpretability, Safety, Robotics, and Geographic Balance

Researcher	Affiliation	Known For	Current Focus
Chris Olah	Anthropic (co-founder)	Mechanistic interpretability pioneer, Distill.pub co-founder	Reverse-engineering model internals for safety
Neel Nanda	Google DeepMind	Created TransformerLens; "A Mathematical Framework for Transformer Circuits"	Leads DeepMind's mechanistic interpretability team; Gemma Scope release
Ajeya Cotra	Open Philanthropy	"Biological anchors" framework for AI-timeline forecasting	AI timelines research, grantmaking strategy
Dan Hendrycks	Center for AI Safety	Created MMLU; co-created Humanity's Last Exam (covered in this site's AGI Benchmarks article)	AI safety benchmarking, x-risk policy advocacy
Stuart Russell	UC Berkeley, CHAI founder	Co-authored the standard AI textbook; "Human Compatible" (2019)	Value alignment and provably-beneficial AI research
Sergey Levine	UC Berkeley; Physical Intelligence co-founder	Pioneered deep RL for robotics	Foundation models for general-purpose robots
Chelsea Finn	Stanford; Physical Intelligence co-founder	Created MAML (model-agnostic meta-learning)	Scalable robot learning toward general behavior
David Silver	Reportedly departed DeepMind, Jan 2026	Led AlphaGo and AlphaZero	Reportedly CEO of new startup "Ineffable Intelligence" — recent, verify directly
Richard Sutton	University of Alberta	Co-authored the foundational RL textbook; "The Bitter Lesson" essay; 2024 Turing Award (confirmed)	Reportedly launched "Oak Lab" for continuously-learning agents
Jürgen Schmidhuber	KAUST	Co-invented LSTM; has long publicly disputed deep-learning credit narratives	AI Initiative Director and Center of Excellence co-chair, KAUST
Bernhard Schölkopf	Max Planck Institute for Intelligent Systems	Foundational kernel methods/statistical learning theory work	Causal representation learning research program
Balaraman Ravindran	IIT Madras	30+ years in reinforcement learning research	Chaired India's AI Governance Guidelines committee; responsible AI research leadership
Jie Tang	Tsinghua University	Chinese academic AI/graph-mining research leadership	Specific current projects need dedicated verification
Jason Wei	Reportedly Meta Superintelligence Labs	Lead author on chain-of-thought prompting (2022); contributed to OpenAI's o1	Employer move unconfirmed — verify before citing
Quoc Le	Reportedly departed Google, 2026	Co-invented seq2seq (2014); led Google's AutoML/neural architecture search	Reported new venture unconfirmed — verify before citing

Addendum

Executives and Others Worth Naming Alongside the 50

This atlas skewed toward technical and research contributors over pure executives, which left a few conspicuous gaps — people who are either founders of labs this site covers or infrastructure figures whose absence undercuts the atlas's own completeness. Added here rather than folded into the tiers above, since their inclusion criteria (leadership and infrastructure impact rather than research authorship) differ from the rest of this piece.

Mustafa SuleymanDeepMind's third co-founder

Co-founded DeepMind in 2010 alongside Demis Hassabis and Shane Legg — the omission of DeepMind's third founder from this atlas was a clear gap. Left DeepMind for Google, then co-founded Inflection AI (2022), then joined Microsoft as CEO of Microsoft AI (2024) following Microsoft's Inflection talent deal. Directly relevant to the "Microsoft AI" lab this site has not yet profiled in a dedicated lineage article.

Sam Altman & Greg BrockmanOpenAI's CEO and President

Every other lab covered in this site's lineage series has its CEO/founder profiled somewhere in this atlas (Hassabis, the Amodeis, Mensch) except OpenAI's Altman and Brockman — an omission given how central both are to the GPT lineage's corporate history, including the November 2023 board crisis. Both remain in their roles as of the most recent available reporting.

Alexandr WangScale AI founder, now Meta's Chief AI Officer

Founded Scale AI, then joined Meta as Chief AI Officer following Meta's ~$14.3 billion investment in Scale AI (2025), leading Meta Superintelligence Labs' TBD Lab — covered in this site's Meta AI (Llama) lineage article's reorganization section, but not previously profiled as an individual here.

Jeff Dean & Sanjay GhemawatGoogle's most senior infrastructure researchers

Dean co-founded Google Brain, co-created MapReduce and BigTable, and led much of Google's large-scale systems and later AI infrastructure work; Ghemawat is his longtime infrastructure collaborator. Both are referenced elsewhere in this article only through an unverified 2026 mass-departure claim (see the Tier 2 entry on Oriol Vinyals) — on their own technical merits, independent of that unconfirmed claim, both belong in any serious accounting of the infrastructure work underlying Google's AI research.

Andrew BartoCo-author of the RL textbook

Co-authored the foundational reinforcement learning textbook with Richard Sutton and shared the 2024 Turing Award with him — an omission given Sutton's own inclusion in Tier 3 credits work Barto co-created.

Also Worth Flagging: Missing Labs This atlas's researcher gaps mirror lab-level gaps in this site's broader Model Case Study series. Alibaba's Qwen team, Amazon's Nova/Titan effort (led in part by former Adept AI leadership), Microsoft AI under Suleyman, and Cohere (co-founded by Transformer author Aidan Gomez, profiled above) do not yet have dedicated lineage articles on this site.

Patterns Across the 50

Three Things This List Makes Visible

Pattern 1 — A Small Number of Advisor Trees Explain a Lot Hinton alone supervised or co-supervised Sutskever (directly) and shaped the intellectual lineage behind Karpathy, Radford, and dozens of others one or two hops removed. Bengio supervised Goodfellow. A handful of PhD advisor relationships from Toronto, Montreal, Berkeley, and Stanford account for a disproportionate share of this entire atlas — a genuinely small set of academic lineages produced most of the field's founding generation.

Pattern 2 — 2025-2026 Has Been an Unusually Volatile Year for Talent This research pass found more contested, single-sourced, or actively-in-motion employer changes than any prior research effort on this site — Barret Zoph alone reportedly changed employers four times in under two years, and several other names carry live, unresolved reporting about their current affiliation. This volatility is itself a data point about the field's current state: talent is moving between a small number of frontier labs and well-funded new ventures fast enough that "current employer" is a genuinely unstable fact for a meaningful fraction of the field's most senior people.

Pattern 3 — Geographic and Institutional Concentration Remains Real Despite deliberate effort to include researchers from India, China, and continental Europe in Tier 3, the overwhelming majority of this atlas's most influential figures trained or work in a small number of US/UK/Canadian institutions (Toronto, Berkeley, Stanford, Cambridge, MIT) or a handful of labs (Google Brain/DeepMind, OpenAI, Meta FAIR). This mirrors the same US/China research-output asymmetry this site's Stanford AI Index companion article documented at the institutional level — visible here at the individual-researcher level too.

Cross-Reference Index

Where Else This Site Covers These People

Researcher	Covered in Depth In
Dario Amodei, Daniela Amodei, Jared Kaplan	The Claude Lineage
Demis Hassabis, Shane Legg	The Gemini Lineage, DeepMind Founding Merger
Arthur Mensch, Guillaume Lample, Timothée Lacroix	The Mistral Lineage
Yann LeCun	The Meta AI (Llama) Lineage
Liang Wenfeng	The DeepSeek Lineage
Alec Radford, John Schulman, Ilya Sutskever	The GPT Lineage, GPT-1: How It Was Built
François Chollet, Dan Hendrycks, Shane Legg	AGI Benchmarks
Fei-Fei Li (via Stanford HAI)	Stanford's 2026 AI Index

⚠️ What's Missing or Uncertain

This atlas contains an unusually high number of flagged, single-sourced 2025-2026 claims, and readers should treat every "Unverified" tag as a genuine research gap, not a formality. Several claims — Ilya Sutskever's SSI product timeline, Andrej Karpathy's reported Anthropic move, Ian Goodfellow's reported venture fund, Barret Zoph's four-employer path, and especially the reported Vinyals/Dean/Ghemawat/Le mass departure from Google and Noam Brown's "Millennium Prize" claim — rest on secondary or aggregator reporting that could not be corroborated against primary sources (official company announcements, the individuals' own statements) during research for this article. This list is also not exhaustive: reasonable readers could substitute several names, and important researchers in China, India, and continental Europe beyond those named in Tier 3 were likely under-covered relative to their actual influence, reflecting a research-access and language-coverage limitation rather than a judgment about their importance.

🔗 Reference Links

🎥 Recommended Videos

But What Is a Neural Network?

3Blue1Brown — the ideas this atlas's researchers spent careers building on

Attention Is All You Need — Paper Explained

Yannic Kilcher — the paper behind this article's eight-author section

🧭 Closing — A Small Cast, Constantly Reshuffling

🎯 The Bottom Line

The number of people who have genuinely shaped this field's technical trajectory is smaller than the size of the industry built on top of their work would suggest — a few dozen advisor-tree-connected researchers, a handful of landmark papers, and a wave of talent now cycling rapidly between a small number of frontier labs and well-funded new ventures. The volatility documented in this atlas is not noise — it is itself a signal about how concentrated, well-resourced, and fast-moving the current phase of the field has become. Six months from now, several of the roles listed here will likely have changed again; the technical contributions behind each name will not.