Shujia Liang (Nathan)

Core Contributor, OpenRSI Foundation

I am a core contributor at OpenRSI Foundation, a non-profit aiming towards open-source infrastructure of recursive self-improvement. My interests span AI, computational physics, statistics, chemistry, history, blockchain, and political economy.

I am currently contributing to Agent's Last Exam V2 (Benchmark), OpenRSI-Index (Artifacts), and OpenRSI-X-Index (Coming soon). If you are interested in contributing any scientific tasks, please contact me.

Previously, I was a PhD student in Computational Physics at Stanford and hold a BS with high honors from UC Berkeley. I co-authored an open standard for zero-knowledge verification of AI agent payments (ERC-8150), won 1st place at ETHDenver 2026 with VAEB, a guardrail for AI agent actions, shipped AI infrastructure serving 600k MAU, and have red-teamed 400+ systems in security reviews.

I am passionate about:

computational science + AI to expand human-knowledge boundaries faster
humanitarian science + AI for deeper comprehension of the past and sharper vision of the future
consensus-based guardrails for AI safety and alignment

GitHub LinkedIn Twitter / X Google Scholar

News

[Mar 2026] Hosted a white paper reading club on the future of autonomous AI agents.

[Feb 2026] Talk at BASS (Denver) on AI agent security: verifiable execution of agent actions and guardrails on agent behavior.

[Jan 2026] Talk at The Fintech Retreat on privacy-preserving selective disclosure for humans and AI agents.

Projects

VAEB — Verifiable Agent Execution Bundle

Jenny Zhang, Justin Cheng, Harry Lee, Shujia Liang (Nathan)

Guardrail for AI agent actions: a zero-knowledge proof checks that each action matches what the user signed off on before it executes

[code]

Custom LAMMPS for Non-Equilibrium Molecular Dynamics

Shujia Liang