Who is Aojie (Justin) Yuan?

Aojie (Justin) Yuan is an honors M.S. student in Computer Engineering at the University of Southern California, researching at USC FORTIS Lab with Yue Zhao, and a Shanghai Jiao Tong University alumnus. He works on agent safety, LLM reasoning, and world models, has 15 papers in 2026 — including NeurIPS 2026 (When Simulation Lies) and Findings of EMNLP 2026 (WeClawArena) — and builds the open-source projects AEGIS and Sovereign-OS.

Are Aojie Yuan, Justin Yuan, and 袁傲杰 the same person?

Yes. He publishes as Aojie Yuan; Justin is his English name and 袁傲杰 is his Chinese name. Google Scholar lists him as "Justin Yuan" (profile bbx7LoIAAAAJ).

What is AEGIS?

AEGIS is an open-source pre-execution firewall and audit layer for AI agents: it checks every tool call before it runs — allow, block, or route to a human — and writes a tamper-evident audit trail, with zero changes to agent code. USC Viterbi News (August 2026) reported that the project is led by Aojie (Justin) Yuan.

What does he research?

Making LLMs reliable enough to trust with real-world resources: auditing and securing AI agents (AEGIS, Auditable Agents, WeClawArena), how models reason internally (hidden error awareness in chain-of-thought, format-agnostic reasoning subspaces), and agent memory and evaluation (AgentIR, When Simulation Lies, CUA-SWE).

How can I contact him?

Email aojieyua@usc.edu, or find him on GitHub (Justin0504) and Google Scholar.