Current
Research scientist at sakana.ai 🗼
Research
I'm interested in how local relates to global, and what that means for theories of intelligence. I take inspiration from algebraic topology, distributed optimization, and neuro-AI. Pragmatically, I'm most interested in multi-agent coordination and local learning alternatives to backprop.
Previous
- 2019–2023Research scientist at Meta
- 2017–2019Lead scientist at CTRL-labs (acquired by Meta)
- 2011–2017PhD in Theoretical Neuroscience, Columbia University Neurotheory
Papers
Recent papers worth reading:
- 2026 Augmented Lagrangian Predictive Coding arXiv
- 2026 Learning Multi-Agent Coordination via Sheaf-ADMM arXiv blog GitHub
- 2025 Sheaf Cohomology of Linear Predictive Coding Networks arXiv GitHub
Full publication list at Google Scholar.
Contact
jeffrey [at] jsseely.com