I build reliable AI agents, tools, and workflows.
- Human-directed agent systems: consent-first research, bounded iterative improvement, and transparent decision workflows
- Repository intelligence and safe code operations: code-aware retrieval, citations, approvals, sandboxed execution, and replayable artifacts
- Deployable AI products and perception data loops: GEO workflows, production-ready applications, and VLM-assisted hard-example mining
| Project | What it demonstrates | Status |
|---|---|---|
| Ya | Consent-first personal research and decision CLI agent featuring user-approved SEA recursive improvement, an ICM curiosity loop for bounded evidence gaps, and a bounded Tree of Agents (ToA) multi-agent architecture | v0.1.6 |
| RepoAirlock | Safety-oriented execution harness for coding agents: Docker isolation, policy enforcement, audit trails, patch replay, HTML reports | v0.1.1-alpha |
| RepoRAG | Source-grounded repository RAG plus an approval-gated code maintenance workbench with LangGraph, MCP, and security guards | v0.1.0-alpha |
| GEO Lens | Production-style full-stack AI SaaS for GEO / AI-search visibility workflows | Live Demo |
- DMS-MultiTask — lightweight driver monitoring model with 1.14M parameters and TensorRT-ready deployment
- VLM Data Closedloop — VLM-assisted hard-example mining, annotation, and augmentation pipeline for cabin monitoring
- Huazhou Dictionary — offline WeChat mini program for Huazhou dialect lookup with 4,310 entries