AI Researcher

Nathan Lambert Portfolio

RLHF and post-training researcher who was a post-training lead at the Allen Institute for AI (Ai2) from 2023 to 2026 and first author of the Tülu 3 report, after leading the RLHF team at Hugging Face; writes the Interconnects newsletter (84,000+ subscribers, per Substack) and finished the free "RLHF Book," also published by Manning in 2026. His site now says he is "doing something new."

RLHFPost-TrainingReinforcement LearningOpen Models

Location

Somewhere on planet Earth, most likely near a laptop.

Companies

Allen Institute for AI (Ai2)Hugging Face

Credentials

  • Ph.D. in Electrical Engineering and Computer Science, UC Berkeley (2022)
  • B.S. in Electrical and Computer Engineering, Cornell University (2017)

Skills

RLHFPost-TrainingReinforcement LearningOpen Models

Notable

  • ACL 2024 Best Theme Paper Award (OLMo) and Best Resource Paper Award (Dolma), per his CV
  • CVPR 2025 Best Paper Honorable Mention (Molmo & PixMo), per his CV
  • Author of "Reinforcement Learning from Human Feedback" (Manning, 2026), free at rlhfbook.com
  • First author of "Tülu 3: Pushing Frontiers in Open Language Model Post-Training" (1,400+ Google Scholar citations)

What makes it work

A breakdown of the choices that make this portfolio stand out.

1

A full open textbook, not just blog posts

The RLHF Book compiles the mechanics of reinforcement learning from human feedback into one structured, freely accessible reference — a materially bigger undertaking than a series of standalone posts, with code for its core chapters, a companion post-training course, and a Manning print edition.

2

Commentary paired with a podcast for primary-source interviews

Interconnects Interviews lets him bring other researchers on directly rather than only characterizing their work secondhand, adding a reporting layer on top of the analysis.

3

Willingness to write candidly about the AI job market and lab dynamics

Posts on the AI research job market and a trip report from inside China's AI labs are riskier to publish than pure technical content, and that candor is part of the newsletter's pitch: the view "from inside the frontier AI labs, minus the hype."

What AI Researchers can take from this

Specific, actionable tips to apply to your own portfolio — no generic advice.

  • If you understand a subfield deeply enough, write the textbook for it — a structured reference outlasts any single blog post.

  • Add a podcast or interview format alongside your writing so primary sources speak in their own words, not just through your summary.

  • Don't be afraid to write candidly about the state of your industry alongside pure technical content — it's part of what makes a research newsletter get read inside the field, not just outside it.

Projects

A textbook, a newsletter, a podcast, and a dashboard all orbit one topic here, as if Nathan Lambert decided post-training deserved its own small media company.

ResearchadvancedMulti-year book project

RLHF Book

RLHFReward ModelingPost-TrainingPython

A free, structured introduction to RLHF and LLM post-training: instruction tuning, reward modeling, RL, direct alignment algorithms, rejection sampling, preference data, and evaluation, with code for the core chapters.

Writing the reference for a subfield shows mastery at a depth individual papers can't.

  • Available as web, PDF, EPUB, and in print

  • Chapters on reasoning and tool use as well as classic RLHF

  • Open source on GitHub

ResearchadvancedOngoing publication

Interconnects

Technical WritingOpen ModelsPodcasting

A Substack newsletter on AI research and open models, with an Interconnects Interviews podcast series alongside it.

Shows a practitioner who can explain frontier work to engineers, researchers, and investors at once.

  • Interview series with other researchers

  • Trip report from inside China's AI labs

  • Posts on the AI research job market

ResearchintermediateOngoing project

ATOM Project

Open ModelsEcosystem Measurement

An effort on open models, listed on his homepage with the Artifacts Hub and an Adoption Dashboard; its companion paper is "The ATOM Report: Measuring the Open Language Model Ecosystem" (2026, with Florian Brand).

Measuring the open-model ecosystem directly, rather than only commenting on it.

  • Launched 2025

  • Paired with an adoption dashboard

  • Co-authored report

Ready to build your portfolio?

Follow the AI Researcher roadmap — skills, projects, and timeline to get hired.

AI Researcher Roadmap