# Eric Zelikman

> Eric Zelikman is the co-founder and CEO of humans&, a human-centric frontier AI lab in Redwood City building foundation models designed for communication and collaboration rather than automation. A former early researcher at xAI, he helped shape Grok's pretraining data and scaled reinforcement learning for reasoning across Grok 3 and Grok 4. Before that, as a Stanford PhD candidate, he authored STaR and Quiet-STaR, foundational papers that taught language models to reason in natural language using their own rationales. In January 2026, humans& emerged from stealth with a $480M seed round at a roughly $4.48B valuation, backed by Nvidia, Jeff Bezos, GV, SV Angel and Emerson Collective.

- **Role:** Co-Founder and CEO at humans&
- **Organizations:** humans&, xAI, Stanford University, Microsoft Research, Google Research, Lazard
- **Nationality:** American
- **Education:** Ph.D. candidate, Computer Science (on leave since March 2024), Stanford University, B.S. in Symbolic Systems, with Honors, Stanford University
- **Known for:** Authored STaR (Self-Taught Reasoner), foundational work on bootstrapping reasoning in language models (NeurIPS 2022), Co-authored Quiet-STaR, enabling language models to generate internal rationales before responding (COLM 2024), Created Parsel, a framework for algorithmic reasoning via decomposition (NeurIPS 2023 Spotlight)

## Career timeline

- **2020** — Graduated Stanford with a B.S. in Symbolic Systems (Honors); developed GANs curriculum for DeepLearning.AI
- **2020** — Joined Lazard as a Deep Learning Engineer
- **2021** — Began Ph.D. in Computer Science at Stanford, advised by Nick Haber and Noah Goodman
- **2022** — Published STaR (Self-Taught Reasoner) at NeurIPS, the first paper to train language models to reason via their own rationales
- **2023** — Parsel accepted as a NeurIPS spotlight; research internships at Microsoft Research and Google Research
- **2024** — Joined xAI as Member of Technical Staff; published Quiet-STaR at COLM
- **2024-2025** — At xAI, contributed to Grok 2 pretraining data, scaled RL for reasoning in Grok 3 Thinking, and built agent RL infrastructure for Grok 4
- **2025** — Left xAI to found humans&; gave a TEDAI San Francisco talk, 'AI should empower people, not replace them'
- **2026** — humans& emerged from stealth with a $480M seed round at a ~$4.48B valuation

## Achievements

- Authored STaR (Self-Taught Reasoner), foundational work on bootstrapping reasoning in language models (NeurIPS 2022)
- Co-authored Quiet-STaR, enabling language models to generate internal rationales before responding (COLM 2024)
- Created Parsel, a framework for algorithmic reasoning via decomposition (NeurIPS 2023 Spotlight)
- Key technical contributor to Grok 2, Grok 3 Thinking, and Grok 4 at xAI
- Raised a $480M seed round for humans& at a ~$4.48B valuation, backed by Nvidia, Jeff Bezos, GV, SV Angel and Emerson Collective
- ACL 2023 Best Reviewer Award (top ~1-1.5%); ICLR 2022 Highlighted Reviewer
- Delivered a TEDAI San Francisco talk in October 2025

## Latest updates

- **2026-01** — humans& emerged from stealth with a $480M seed round at a ~$4.48B valuation, backed by Nvidia, Jeff Bezos, GV, SV Angel and Emerson Collective.
- **2025-10** — Delivered a TEDAI San Francisco talk titled 'AI should empower people, not replace them.'
- **2025-09** — Left xAI to co-found and lead humans& as CEO.

## Links

- Website: https://zelikman.me
- LinkedIn: https://www.linkedin.com/in/ericzelikman
- Twitter/X: https://x.com/ericzelikman
- GitHub: https://github.com/ezelikman

---

Profile page: https://yespress.io/eric-zelikman
Published by YesPress — https://yespress.io
Last updated: 2026-06-24