Built from work this engineer published themselves. It makes no claim that they are available for hire, and it is marked noindex until they claim it. Source: lilianweng.github.io. This is you, or it should not be here? Write to hello@aiengineer.me.
ai-engineer
Lilian Weng
Survey-grade write-ups on agents, hallucination, and reward hacking.
Writes long technical surveys that practitioners treat as reference material: LLM-powered autonomous agents, extrinsic hallucination, adversarial attacks on LLMs, and reward hacking. Previously led safety systems research at OpenAI, and the writing keeps that framing — mechanisms, failure modes, and the evidence for each.
Technical stack & depth
Agents
Tool use, multi-step orchestration, and recovery when a step fails.
Guardrails
Policy, refusal, red-teaming, and the cases the model must not invent.
Evaluation
Evals, traces, and a quality bar that is not a vibe check.
Engineered solutions
- Write-upLLM Powered Autonomous Agentslilianweng.github.io
Author of the agent-architecture survey used as a common reference for planning, memory, and tool use.
- Portfoliolilianweng.github.iolilianweng.github.io
Archive of technical surveys on hallucination, adversarial attacks, and reinforcement learning.
Something here out of date or wrong? Write to hello@aiengineer.me and it will be corrected or removed.