Who Is Shengjia Zhao?

Shengjia Zhao is a rising star in artificial intelligence research. He earned his bachelor’s degree from Tsinghua University in 2016, followed by a PhD in Computer Science from Stanford University in 2022.

His doctoral work centered on synthetic data generation, self-supervised learning, and data governance for pretraining—areas that later became central to industrial AI systems.

Key Contributions at OpenAI

Joining OpenAI in 2022, Zhao quickly became integral to the early wave of breakthroughs:

• He co-created ChatGPT and was a major contributor to GPT‑4, including its compact variants like GPT‑4.1 and o3.
• Zhao led OpenAI’s synthetic data research, crucial for efficient training of large-scale language models.
• He played a key role in OpenAI’s post‑training and reasoning mechanisms, particularly in the development of AI reasoning paradigms like chain‑of‑thought models.

His departure was confirmed alongside other major researchers—Shuchao Bi, Jiahui Yu, and Hongyu Ren—as among the top donor losses for OpenAI.

Today: At the Helm of Meta’s Superintelligence Labs

In July 2025, Meta appointed Zhao as Chief Scientist of its newly launched Meta Superintelligence Labs (MSL). He is also recognized as a co‑founder of the lab, alongside leadership figures like Alexandr Wang (Chief AI Officer) and Nat Friedman (applied research).

In his new role, Zhao is tasked with shaping the lab’s scientific agenda, guiding research on foundational models, and working closely with Zuckerberg and Wang to pursue artificial general intelligence (AGI) and beyond.

Meta has consolidated teams including FAIR, LLaMA engineering, and product-focused research into MSL, aiming to establish a hub for cutting-edge AI development.

4. Research Focus & Legacy

A) Synthetic Data & Efficient Pretraining

Zhao’s synthetic data work addresses a persistent bottleneck in model scaling—how to train massive models when high‑quality real-world datasets are limited. His solutions make training more efficient, robust, and privacy-conscious.

B) Reasoning Models & Post‑Training Refinement

Zhao contributed to models facilitating chain-of-thought reasoning, enabling AI to structure complex problem solving—an essential capability for future AGI systems.

C) Multimodal and Mini Model Innovation

He co-developed compact versions of GPT‑4 as well as multimodal reasoning models (combining text, image, perhaps voice), advancing wider application and accessibility.

Industry Recognition and Impact

Zhao’s academic output garners significant citations, with his Google Scholar profile showing high h-index and impact metrics in recent years. In 2022, he was awarded ICLR’s Outstanding Paper Award.

His transition from OpenAI to Meta was a focal point in discussions about the intensifying AI talent war—he was part of a wave of elite AI researchers recruited from top rivals.

OpenAI CEO Sam Altman publicly criticized Meta’s high-value hiring tactics as “crazy,” though several departing researchers denied receiving such enormous packages. Tech leaders, including Google DeepMind CEO Demis Hassabis, recognized Meta’s urgency to catch up in AI, saying their strategy was rational given the gap to close.

Broader Context & What It Means

Meta's Strategic AI Pivot

Zhao’s appointment marks Meta’s bold push into AGI and superintelligence, signaling a shift from social networking into next-generation AI leadership. The company has committed billions, including a $14–15 billion investment in infrastructure like Scale AI to support its ambitions.

Leadership has made explicit its intention to build open-source AI, differentiating itself from closed-source strategies embraced by OpenAI and other big labs.

Vision for AGI & Accessibility

Under Zhao’s scientific direction, MSL aims to develop “personal superintelligence for everyone”—multimodal systems that can reason, synthesize, and assist across domains while emphasizing transparency and broad accessibility.

Talent & Cultural Considerations

Hiring elite researchers is one side — Meta now must build a sustainable culture of innovation and alignment rather than just purchasing talent. Observers question whether huge financial packages alone can foster mission-driven AI research environments.

What’s Next: Projects & Leadership Goals

Future Models & Releases

Meta is reportedly preparing to release a more capable model than LLaMA‑4, under internal codename “Behemoth”, expected later in 2025. Zhao will likely drive research directions for that system and beyond.

Ethical & Safety Alignment

Given his leadership role, Zhao is well-positioned to influence frameworks around AI safety, alignment, and ethics, ensuring next-gen systems are socially responsible and transparent.

Publications & Open Research

Meta has signaled an open-source ethos. With Zhao leading scientific direction, MSL may publish more fundamental research and open policy not typical in closed‑door AI models.