The history of technology is written not just in laboratories and corporate campuses, but in the moments when the people who built those institutions decide to step away and build something of their own. In artificial intelligence, that moment has arrived — and it has arrived spectacularly. The researchers who invented the Transformer architecture, developed AlphaFold, pioneered reinforcement learning, and shipped ChatGPT are now racing to found their own labs, backed by billions of dollars in venture capital and armed with the most valuable commodity in the industry: credibility. What follows is a guide to the researchers who quit — and what they are building next.
1. Yann LeCun — AMI Labs
Yann LeCun is one of the most celebrated figures in the history of artificial intelligence. A Turing Award winner (alongside Geoffrey Hinton and Yoshua Bengio in 2018), LeCun is widely credited as one of the pioneers of convolutional neural networks. He spent 12 years at Meta, serving as founding director of Facebook AI Research (FAIR) and later as chief AI scientist. During this time, he championed open-source AI development and became a prominent, often contrarian voice in the debate about the future of large language models, repeatedly arguing that LLMs alone cannot achieve human-level intelligence.
In November 2025, LeCun announced his departure from Meta to co-found Advanced Machine Intelligence (AMI) Labs, a Paris-based startup focused on building “world models” — AI systems that understand the physical world rather than merely predicting text. He serves as executive chairman, with Alexandre LeBrun (formerly CEO of health-tech startup Nabla) as CEO. In March 2026, AMI Labs announced it had raised $1.03 billion at a $3.5 billion pre-money valuation, co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions, with high-profile angels including Eric Schmidt and Mark Cuban. The company is pursuing LeCun’s Joint Embedding Predictive Architecture (JEPA) as an alternative foundation for next-generation AI.
2. Fei-Fei Li — World Labs
Fei-Fei Li is frequently called the “godmother of AI” — a title she earned through her creation of ImageNet, the landmark dataset that catalyzed the deep learning revolution in computer vision. A professor at Stanford University and co-director of Stanford’s Human-Centered AI Institute, Li previously served as chief scientist of AI/ML at Google Cloud from 2017 to 2018. Her work on making AI development more inclusive and safety-conscious has made her one of the field’s most respected voices outside of pure research.
In April 2024, Li founded World Labs, a startup developing “large world models” — AI systems that reason about and operate within three-dimensional environments. The company went from founding to unicorn in just four months, raising $230 million at a $1 billion valuation, backed by Andreessen Horowitz, NEA, Nvidia, and others. In late 2025, World Labs launched its first product, Marble, a platform that lets users create, edit, and share high-fidelity, persistent 3D worlds. By February 2026, the company had raised an additional $1 billion led by Autodesk, bringing its total raised to over $1.2 billion, with discussions reportedly underway for a new round at a $5 billion valuation — a fivefold increase in roughly a year.
3. Ilya Sutskever — Safe Superintelligence (SSI)
Ilya Sutskever is one of the most important figures in modern AI, and his departure from OpenAI sent shockwaves through the industry. A co-founder of OpenAI itself, Sutskever served as its chief scientist and was a central architect of many of the company’s defining breakthroughs, including the GPT series of models. His academic work on deep learning — including the seminal 2012 AlexNet paper with Geoffrey Hinton — helped spark the current era of neural network-based AI. In late 2023, Sutskever played a significant role in OpenAI’s boardroom drama, voting to temporarily remove CEO Sam Altman, before later reconciling with the decision and ultimately leaving the company in May 2024.
In June 2024, Sutskever co-founded Safe Superintelligence Inc. (SSI) alongside Daniel Gross, former head of AI at Apple, and Daniel Levy, an AI researcher. The company’s stated mission is singular and audacious: build one product, a “safe superintelligence,” with no commercial distractions along the way. Operating with extraordinary secrecy from offices in Palo Alto and Tel Aviv, SSI has quickly become one of the most valuable pre-product startups in history. In April 2025, the company raised $2 billion in a round led by Greenoaks Capital at a $32 billion valuation — a sixfold increase from its $5 billion valuation just eight months earlier — with backing from Andreessen Horowitz, Lightspeed, Alphabet, and Nvidia.
4. Mira Murati — Thinking Machines Lab
Mira Murati rose to global prominence during OpenAI’s dramatic leadership crisis of November 2023, when she served as interim CEO following the brief ouster of Sam Altman. As OpenAI’s chief technology officer, she oversaw the development of some of the company’s most iconic products, including ChatGPT, DALL-E, and voice mode. After more than six years at the company, Murati departed OpenAI in September 2024, citing concerns about the company’s strategic direction. She quickly assembled a team that reads like an OpenAI alumni reunion: co-founders include reinforcement learning pioneer John Schulman, former VP of research Barrett Zoph, and alignment researcher Lilian Weng.
In February 2025, Murati launched Thinking Machines Lab, describing its mission as building multimodal AI that is more customizable, capable, and safer. The company’s first product, Tinker, launched in October 2025, enabling companies and researchers to fine-tune AI models for specific industries with far less complexity than traditional training methods. The startup raised $2 billion in one of the largest seed rounds in Silicon Valley history at a $12 billion valuation, led by Andreessen Horowitz with participation from Nvidia, Accel, ServiceNow, Cisco, AMD, and Jane Street. By November 2025, Thinking Machines was reportedly in advanced talks to raise a fresh round at a valuation of $50 billion, reflecting extraordinary investor confidence in Murati’s team.
5. Mustafa Suleyman — Inflection AI (now Microsoft AI)
Mustafa Suleyman co-founded DeepMind in 2010 — the London AI lab later acquired by Google for approximately £400 million — and helped build it into one of the world’s most influential AI research organizations. Born in North London to a Syrian father and English mother, Suleyman dropped out of Oxford to co-found the Muslim Youth Helpline before pivoting to technology. After leaving DeepMind in 2019, he served as VP of AI products and policy at Google before co-founding Inflection AI in 2022 alongside LinkedIn co-founder Reid Hoffman and researcher Karen Simonyan. He is also the author of The Coming Wave, described by Bill Gates as his “favourite book about AI.”
Inflection AI raised over $1.3 billion in funding and launched Pi, a conversational AI assistant designed to be emotionally intelligent and empathetic. The company positioned itself as a more human-centered alternative to ChatGPT. In March 2024, Microsoft appointed Suleyman as CEO of its newly created Microsoft AI division, which oversees consumer AI products including Copilot, Bing, and Edge. In a deal widely described as an effective acqui-hire, Suleyman and key Inflection team members joined Microsoft. Inflection itself pivoted to enterprise AI software. At Microsoft, Suleyman has articulated a vision of “Human Superintelligence” — AI that remains firmly in service of people — and has positioned the company for greater independence from OpenAI in the development of frontier models.
6. Arthur Mensch — Mistral AI
Arthur Mensch, born in 1992 near Paris, studied at the prestigious École Polytechnique and earned a PhD in machine learning from Université Paris-Saclay. He spent three years as a senior research scientist at Google DeepMind in Paris, where he contributed to high-impact projects including the multimodal Flamingo model and early Gemini research. In April 2023, he co-founded Mistral AI alongside Guillaume Lample and Timothée Lacroix — two colleagues who had worked on large-scale language models at Meta — after the trio recognized an opening to build an efficient, open-source European alternative to the American AI giants.
Mistral AI quickly established itself as Europe’s AI champion and the most credible rival to OpenAI and Anthropic outside the United States. The company released a series of models — Mistral 7B, Mixtral 8x7B, and others — that punched well above their weight, often matching larger proprietary models at a fraction of the compute cost. Mistral’s consumer chatbot, Le Chat, launched on mobile in early 2025 and gained millions of users. In September 2025, the company completed a €2 billion Series C led by Dutch chip equipment maker ASML at a €11.7 billion ($14 billion) valuation, making Mensch and his co-founders France’s first AI billionaires, each with an estimated net worth of approximately $1.1 billion.
7. Simon Kohl — Latent Labs
Dr. Simon Kohl is a computer scientist and AI researcher who spent his career at the intersection of machine learning and biology. He joined Google DeepMind as an intern and rose to become a senior research scientist and co-lead of the protein design team, playing a key role in building AlphaFold2 — the Nobel Prize-winning AI system that predicted the 3D structure of over 200 million proteins and is widely considered one of the greatest scientific achievements of the decade. After AlphaFold2, Kohl worked on generative protein design, growing frustrated that the technology’s commercial applications were not being fully pursued within DeepMind’s structure.
In 2023, Kohl founded Latent Labs, a London- and San Francisco-based AI biotech startup with the mission of making biology programmable. The company uses generative AI to design new protein-based therapeutic molecules, such as antibodies and enzymes, from scratch — a capability that was essentially impossible before the AlphaFold era. In February 2025, Latent Labs emerged from stealth with $50 million in funding, including a $40 million Series A co-led by Radical Ventures and Sofinnova Partners, with notable angel investors including Google chief scientist Jeff Dean and Cohere founder Aidan Gomez. The company is now partnering with major biopharma firms to accelerate their drug discovery programs using Latent’s generative biology platform.
8. David Silver — Ineffable Intelligence
David Silver is one of the most accomplished reinforcement learning researchers in the world, and one of the longest-serving members of the original DeepMind team, having joined the company at its founding in 2010. He led the research teams behind AlphaGo, AlphaZero, and MuZero — AI systems that successively defeated the world’s best human players at Go, chess, and shogi, stunning the scientific community with their capability. Silver is also a professor at University College London and a recipient of numerous prestigious awards for his contributions to RL. His work fundamentally shaped how the field thinks about intelligence emerging from learning rather than explicit programming.
In late 2025, Silver quietly left Google DeepMind after years during which he had reportedly been on sabbatical, never formally returning to the company. In January 2026, company filings with the UK’s Companies House confirmed he had been appointed a director of Ineffable Intelligence, a new AI startup based in London. The company is in the active process of recruiting AI researchers and seeking venture capital funding. Silver has told friends he wants to return to “the awe and wonder of solving the hardest problems in AI,” and sees superintelligence as the defining unsolved challenge in the field. Specific details about Ineffable Intelligence’s research direction and funding status have not yet been publicly disclosed.
9. Misha Laskin & Ioannis Antonoglou — Reflection AI
Misha Laskin and Ioannis Antonoglou share a deep history at Google DeepMind that began with one of AI’s most celebrated moments. Antonoglou, born in Thessaloniki, Greece, was DeepMind employee #25 and researcher #6, joining in 2012 and becoming a core architect of AlphaGo — the system that defeated Go world champion Lee Sedol in 2016. Laskin, a former quantum physics PhD, read the AlphaGo paper and changed his career trajectory, eventually joining DeepMind where he led reward modeling for the Gemini project. The two collaborated on Gemini 1 and 1.5 before deciding to strike out on their own.
In March 2024, Laskin and Antonoglou co-founded Reflection AI, a New York-based startup applying reinforcement learning to autonomous software engineering. The company’s flagship product, Asimov, is an AI agent that can ingest entire codebases and autonomously plan, write, and review code. Reflection launched in March 2025 with $130 million in initial funding from Sequoia, CRV, and Lightspeed. By October 2025, the company had raised an additional $2 billion at an $8 billion valuation — one of the largest rounds ever for an open-source AI company — backed by Nvidia, Eric Schmidt, DST, Citi, and others. Reflection has positioned itself as America’s answer to China’s DeepSeek, committing to release model weights publicly while keeping training pipelines proprietary.
10. Jonathan Godwin — Orbital Materials
Jonathan Godwin is a computer scientist and former senior research engineer at Google DeepMind, where he spent approximately five years working on AI for scientific discovery. He led a team applying large-scale machine learning to materials science — specifically simulating atomic-scale interactions to predict and design new materials. His work at DeepMind overlapped with the era of AlphaFold and showed him the immense untapped potential of applying these techniques to the physical world. Crucially, Godwin felt that Google was not structured to translate computational predictions into actual physical materials — a gap he was determined to fill himself.
In 2022, Godwin co-founded Orbital Materials in London, alongside James Gin-Pollock and Daniel Miodovnik. The startup uses generative AI to design advanced materials for applications including carbon capture, energy storage, and sustainable aviation fuel. Orbital’s model, Linus, takes natural language prompts and generates 3D molecular structures optimized for specific properties. The company has raised approximately $21 million in funding, including a $16 million Series A led by Radical Ventures with participation from Toyota Ventures and the Sequoia Scout Fund. Orbital has also open-sourced Orb, a state-of-the-art simulation model, and established a wet lab in New Jersey where AI-generated material predictions are physically validated.
11. Liam Fedus — Periodic Labs
Liam Fedus — full name William Fedus — is one of the most respected researchers to have emerged from OpenAI’s storied post-training team. He joined OpenAI in 2022 after leaving Google, where he had worked on language model research at Google Brain. At OpenAI, Fedus was part of the small team that created ChatGPT and rose to become Vice President of Research for Post-Training, overseeing the refinement processes that shaped the behavior of models like GPT-4 and its successors. He was promoted to VP in late 2024, even as a wave of senior departures swept through the company.
In March 2025, Fedus announced his departure from OpenAI to found a materials science AI startup, later named Periodic Labs, alongside former Google Brain colleague Ekin Dogus Cubuk. Cubuk had co-authored the landmark GNoME paper on AI-driven crystal discovery and helped build a fully automated robotic lab at Google. Their shared insight: advances in reasoning models (Fedus’s speciality) combined with scientific simulation (Cubuk’s domain) had finally made AI-automated materials discovery viable. OpenAI initially planned to invest in Periodic, though it ultimately did not participate. The company came out of stealth in October 2025 with a massive $300 million seed round led by Andreessen Horowitz, with Felicis, Bezos Expeditions, Eric Schmidt, and others as backers, valuing the company at approximately $1 billion.
12. Rhythm Garg, Linden Li & Yash Patil — Applied Compute
Rhythm Garg, Linden Li, and Yash Patil represent a new generation of AI founders who built their credentials through direct work on frontier models at OpenAI. All three are Stanford University graduates who joined OpenAI shortly after completing their degrees and worked on some of the company’s most strategically significant projects. Garg was a core contributor to the development of OpenAI’s o1 reasoning model, the company’s first system trained explicitly with reinforcement learning to improve multi-step reasoning. Patil was a key member of the agentic Codex team, focused on coding AI capabilities, while Li built the machine learning infrastructure for reinforcement learning training.
In May 2025, the trio founded Applied Compute, a startup building what they call “Specific Intelligence” — custom AI agents trained on a company’s proprietary data rather than general-purpose public datasets. The idea is that enterprises’ real competitive advantage lies not in access to general models, but in AI systems shaped by their unique institutional knowledge. Applied Compute raised $20 million at a $100 million valuation in June 2025 in a round led by Benchmark, with Sequoia also participating. By October 2025, the company had raised $80 million in a Series A at approximately a $700 million valuation, with clients including DoorDash, Mercor, and AI engineering firm Cognition. By January 2026, the company was in talks to raise its next round at a $1.3 billion valuation, less than a year after launching.
13. Shiv Rao — Abridge
Shiv Rao is something of an anomaly in the AI founder landscape: a practicing cardiologist who runs a multi-billion-dollar startup. After training at the University of Pittsburgh Medical Center, where he became an attending cardiologist and faculty member, Rao spent six years working on health data investments at UPMC’s venture arm. His motivation to build Abridge was deeply personal — he was spending his evenings dictating patient notes, a ritually frustrating and time-consuming process he knew could be solved with AI. He co-founded Abridge in 2018 alongside Sandeep Konam and Florian Metze, emerging from the Pittsburgh Health Data Alliance.
Abridge builds AI-powered medical scribing software that records and transcribes doctor-patient conversations in real time, automatically generating structured clinical notes. The company became the dominant AI scribe platform at U.S. health systems, deployed across more than 150 major institutions including Mayo Clinic, Kaiser Permanente, and Yale New Haven Health, all of which also became investors. By early 2025, Abridge raised a $250 million Series D at a $2.75 billion valuation, then doubled its valuation again just four months later with a $300 million Series E at a $5.3 billion valuation, led by Andreessen Horowitz and Khosla Ventures. As of mid-2025, Abridge had an annual recurring revenue run rate of $117 million and commanded roughly 30% market share in ambient clinical documentation in the United States.
14. Noam Shazeer — Character.AI
Noam Shazeer joined Google in 2000 and spent two decades building some of the most consequential systems in the company’s history. He is perhaps best known as a lead co-author of the 2017 paper “Attention Is All You Need,” which introduced the Transformer architecture — the foundational design underpinning ChatGPT, Gemini, Claude, and virtually every modern large language model. At Google, Shazeer and colleague Daniel De Freitas built Meena, a highly capable conversational AI chatbot. When Google declined to release Meena to the public out of safety concerns, Shazeer and De Freitas resigned in 2021 to launch their own venture.
The pair co-founded Character.AI, a platform allowing users to converse with customizable AI characters representing anyone from historical figures to fictional characters. The service gained massive popularity among younger users and reached millions of daily active users by 2023. Character.AI raised $193 million in funding, achieving a peak valuation of approximately $2.5 billion. In August 2024, Google entered into a $2.7 billion licensing and personnel agreement with Character.AI, bringing Shazeer and De Freitas back to Google DeepMind to co-lead the Gemini project — described as one of the most expensive rehire deals in technology history. Shazeer, who owned an estimated 30-40% of Character.AI, netted between $750 million and $1 billion from the transaction.
15. Ashish Vaswani & Niki Parmar — Essential AI
Ashish Vaswani and Niki Parmar are two of the six co-authors of “Attention Is All You Need,” making them among the most directly consequential researchers in the history of modern AI. Both worked on the Google Brain team before co-founding Adept AI in 2021 alongside David Luan, where they worked on training neural networks to perform general software tasks on behalf of users. After leaving Adept in late 2022, Vaswani and Parmar set out to apply their deep expertise in transformer architectures to enterprise AI automation. Their credentials alone were enough to trigger intense investor interest before they had even announced a product.
In 2023, Vaswani and Parmar co-founded Essential AI in San Francisco, with the mission of building an “Enterprise Brain” — a full-stack AI platform that automates time-consuming and repetitive knowledge work within large organizations. The company focuses on making enterprise AI customizable and practically deployable, rather than relying on general-purpose models. Essential AI raised a $8.3 million seed round led by Thrive Capital, followed by a $56.5 million Series A led by March Capital with participation from Google, Nvidia, AMD, and Thrive Capital — bringing its total funding to nearly $65 million. The company, which describes its approach as building AI that learns to “increase productivity by automating monotonous workflows,” has attracted clients across the enterprise technology sector.
16. David Ha — Sakana AI
David Ha has one of the most eclectic backgrounds in AI research: he began his career as a derivatives trader and Managing Director at Goldman Sachs Japan before pivoting to a PhD at the University of Tokyo and joining Google as a research scientist. At Google Brain, he led the Japan team and developed a reputation for exploring nature-inspired approaches to AI — particularly systems that evolve, self-organize, and adapt rather than being trained on brute computational scale. He later became head of research at Stability AI before departing to launch his own lab. His co-founders at Sakana include Llion Jones, co-author of “Attention Is All You Need,” and Ren Ito, a veteran of Japanese tech and venture circles.
In 2023, Ha co-founded Sakana AI in Tokyo, building AI models inspired by natural systems — specifically the idea that smaller models can “breed” and evolve through collective intelligence, dramatically reducing the compute requirements of frontier AI. The name “Sakana” (Japanese for fish) reflects the company’s philosophy of emergent collective intelligence. Sakana quickly became Japan’s fastest startup to reach unicorn status, raising $214 million in a Series A at a $1.5 billion valuation in 2024 from Nvidia, MUFG, KDDI, and others. In November 2025, it raised a further $135 million Series B at a $2.65 billion valuation — the highest ever for an unlisted Japanese startup — with backing from Khosla Ventures, In-Q-Tel (the CIA’s venture arm), and Citigroup. Sakana’s AI Scientist system made history in early 2025 when it became the first AI to have a research paper accepted by a peer-reviewed machine learning conference.