I am a computer science student and systems engineer interested in building reliable AI systems.
Homepage: haochenglu.github.io
- Agent Infra - durable, permission-aware runtimes for long-lived agent workflows.
- Recursive Self-Improvement (RSI) - data-centric training, benchmark design, and evaluation for agents.
- Molecular Parsing - scientific document understanding, molecule detection, and structured chemical extraction from PDFs.
- Context & Runtime Governance - versioned context, pre-action guardrails, and verifiable execution.
- RSIGym - agent-native environment for recursive self-improvement (paper)
- Gödel Forest - co-evolving multi-agent search for data-centric RSI (paper)
- RSIBench-Data - benchmarking LLM agents as data-centric post-training researchers (paper)
- ContextVC - Git-native context control plane for AI coding agents
- Terminal-Bench: rsi-gatekeeper - task grading the acceptance step that recursive self-improvement depends on (open PR)
Incoming M.S. student in Computer Science at Columbia University; B.S. in Computer Science from NYU Shanghai.
I enjoy turning research ideas into reproducible systems, benchmarks, and open-source tools.