Miju Labs

← All buyers · Meta (Meta Superintelligence Labs)

S

Shengjia Zhao

Chief Scientist, Meta Superintelligence Labs · Meta (Meta Superintelligence Labs)

medium confidenceTier A · Talk to firstFrontier labstext · image · video · multimodal

Chief Scientist of Meta Superintelligence Labs, announced by Mark Zuckerberg in July 2025. A co-creator of ChatGPT and co-author of GPT-4 at OpenAI, where he also led synthetic data work. He sets MSL's research agenda, which determines what data the lab buys, but his own centre of gravity is reasoning and text rather than visual taste.

Background

Zhao did his PhD at Stanford with Stefano Ermon after Tsinghua, working on generative models and calibration, then joined OpenAI where he was a co-author on the ChatGPT and GPT-4 papers, worked on o1-mini, and led synthetic-data efforts. He was recruited to Meta in the June-July 2025 hiring wave and named Chief Scientist of Meta Superintelligence Labs on 25 July 2025.

The appointment was not smooth. The Financial Times reported in August 2025 that he had threatened to quit within days of joining and return to OpenAI, and was subsequently given the chief scientist title. Yann LeCun, who remained Meta's Chief AI Scientist at FAIR, publicly clarified the distinction between the roles. That instability is part of the picture: MSL has seen substantial churn since, including the departure of multimodal lead Jiahui Yu in August 2026.

What they run now

Research direction for Meta Superintelligence Labs, working alongside Chief AI Officer Alexandr Wang. MSL's structure is TBD Lab (frontier models), FAIR, Products and Applied Research under Nat Friedman, and MSL Infra. The shipped output under this regime includes Muse Spark (April 2026), Muse Image and Muse Video — so the lab does have a visual-generation line, even though Zhao's own history is text and reasoning.

What they buy

Inference, at the agenda-setting level. His OpenAI background in synthetic data suggests a bias toward generated rather than purchased human data, which cuts against Julian. Meta's known posture — the Scale AI investment and Alexandr Wang's arrival — means data supply is partly internalised. Where external human judgement would still be needed is precisely where synthetic data fails: aesthetic preference, cultural nuance, professional craft. Whether Zhao believes that is unknown.

Career

  1. 2022 – 2025Research scientist, OpenAI — Co-creator of ChatGPT; GPT-4 co-author; o1-mini; synthetic data
  2. 2025-07 – presentChief Scientist, Meta Superintelligence Labs — Announced by Zuckerberg 25 Jul 2025
  3. —Stanford University, PhD, Computer Science
  4. —Tsinghua University, Bachelor's

Interviews & talks

Why it matters here

Medium-low for creative data specifically, despite the seniority. He is the wrong modality and, given the Scale AI relationship, Meta is the least likely of the four labs to buy human data from an outside specialist. His value in the index is as context: he explains why MSL's data decisions look the way they do. Julian should not build an outreach plan around him.

How to reach

Not a target. Speculation: if Meta ever becomes worth approaching on creative data, the route is through the Products and Applied Research side (Nat Friedman) or through whoever now leads multimodal after Jiahui Yu's exit, not through the Chief Scientist.

What we could not establish
  • No verified X handle found.
  • Whether he still holds the role given the reported churn at MSL — most recent confirmation in this pass is from 2025 coverage plus the roster.
  • How MSL sources human data given the Scale AI relationship.
  • His views on human versus synthetic data for subjective judgement are not public.

Why they are on the map

Oversees the technical roadmap for the full Muse series, including Muse Image and Muse Video; described as a synthetic-data leader. source — evidence dated 2026-07

Decides whether the problem is funded.

How Meta (Meta Superintelligence Labs) buys data

Aug 2025: TBD Lab researchers reportedly preferred Surge and Mercor data over Scale's, despite Meta's $14.3B Scale stake. Meta has also partnered with Mercor on RL environments (TechCrunch, Sep 2025). In Aug 2025 Meta licensed Midjourney's 'aesthetic technology' to make its models 'more visually appealing', which is a direct signal of paying for visual taste. Muse Image/Video (Jul 2026) are marketed on human-preference Elo from internal ablations and Arena. SemiAnalysis (Jul 2026) reports about 3,000 engineers working on RL task/environment creation.

Alexandr Wang (Chief AI Officer, org context only) heads MSL. Since Aug 2025 it has four parts: TBD Lab (frontier models), FAIR (Rob Fergus), Products and Applied Research (Nat Friedman) and MSL Infra (Aparna Ramani, Amir Frenkel). Shengjia Zhao is Chief Scientist and owns the Muse roadmap; Jiahui Yu leads multimodal, including Muse Image/Video. No human-data owner is public; data-vendor decisions appear to sit with TBD Lab researchers and ex-Scale staff.

Known vendors: scale-ai, surge-ai, mercor, lmarena, Midjourney (aesthetics tech licence)

Sources

  1. Meta names Shengjia Zhao as chief scientist of AI superintelligence unit
  2. Meta announces its Superintelligence Labs Chief Scientist
  3. FT via Techmeme: Zhao threatened to quit within days of joining
  4. Shengjia Zhao, LinkedIn