ByteDance AI Researcher Gu Quanquan Leaves to Pursue Startup Dreams
ByteDance Loses AI Talent as Gu Quanquan Departs
Gu Quanquan, one of ByteDance's top artificial intelligence researchers, has confirmed his resignation from the company's Seed team. The news comes as China's tech giant accelerates commercialization of its AI business units.
In a social media post marking his last day, Gu reflected on his three-year journey developing cutting-edge AI applications in drug discovery and language model training. His departure has sparked intense speculation about where this sought-after researcher might land next.
From Molecules to Machine Learning
Gu's most notable achievement was leading development of SeedFold, ByteDance's breakthrough bio-molecular structure prediction model. Remarkably, this system outperformed DeepMind's renowned AlphaFold3 in multiple public benchmark tests. He also pioneered innovations including:
- The protein binder design platform SeedProteo
- The DPLIV series of protein language models
- Scalable infrastructure for training large language models including Seed2.0
"When we first saw SeedFold's results compared to AlphaFold, that's when we knew ByteDance had something special," shared a researcher familiar with both systems, speaking on condition of anonymity.
Shifting Priorities
In early 2025, Gu pivoted to focus on large language model pretraining, establishing ByteDance's LLM optimization team. However, the company downplayed his role in a statement to media:
"While Mr. Gu contributed to important projects, he was not leading our AI drug development or pretraining business units," a ByteDance representative clarified.
What's Next for the AI Visionary?
The big question now: Where will Gu take his expertise next? Industry insiders overwhelmingly predict he'll join the booming AI4S (AI for Science) startup scene. Multiple sources confirm he's already drawn interest from top-tier venture capital firms eager to back his next venture.
Key Points:
- Gu Quanquan departs ByteDance after three years leading AI drug discovery projects
- His SeedFold molecular model outperformed DeepMind's AlphaFold3
- Shifted focus to large language model training in 2025
- Expected to launch an AI-for-science startup in drug development
- Already attracting significant investor attention