Ph.D. Student at Tsinghua University, working at the intersection of urban science and artificial intelligence.
Intern at ByteDance, working on post-training, agentic reinforcement learning, and data synthesis.
I build building-scale urban datasets and simulation methods, evaluate and align generative models for scientific discovery, and turn AI-agent research into open-source systems people can use.
50+ public projects · 1.2k+ GitHub stars · 32M+ buildings mapped · 45K+ dataset downloads
| Area | Project | What It Does |
|---|---|---|
| AI products | MemoMind | Persistent local memory for AI agents, with an evolving knowledge graph across sessions |
| HumanSkillTree | AI-powered lifelong-learning system grounded in cognitive and learning science | |
| Zync | IDE for orchestrating Claude Code, Codex, Aider, and other coding agents in parallel | |
| DigitalMe | Control coding agents from Feishu, Telegram, or Slack with monitoring and alerts | |
| SmartBidding 🔒 | Production engineering-bid intelligence platform spanning OCR, structured review, retrieval, generation, and audit | |
| AI infrastructure | AI API Gateway 🔒 | Private model-routing, shared-service access, device isolation, and credential-control architecture |
| Data infrastructure | GEHI Batch Router | Restartable historical-imagery routing with actual-date provenance, quality gates, and auditable JSONL manifests |
| DayLife | Local-first personal activity understanding and AI-assisted reflection |
This is a selected portfolio. 🔒 Private repositories are listed for portfolio context; source access is restricted. Explore the complete public project catalog for more research code, agent skills, developer infrastructure, and applications.
Supporting systems: InfoHub · web-novel-crawler · HomeSync · Desktop Controller · Recall · Prompt Harness · Experiment Log · Note2Video · AI Reviewer · CiteMind · k2md
These projects cover the supporting information, agent, knowledge, and data-collection work around the main systems.
| Work | Contribution | Links |
|---|---|---|
| Building-Scale Urban Analytics and Simulation | A research vision connecting building data foundations, generative AI, and bottom-up urban simulation | Paper |
| AI4US | 1M+ synthesis trials revealing how GenAI captures urban science while oversimplifying population complexity | Code · arXiv |
| UrbanAlign | Training-free VLM–human preference alignment with interpretable semantic calibration | Code · arXiv |
| CMAB → CMAB-T | A nationwide multi-attribute building dataset extended into long-term building-vector tracking | Paper · Dataset · History code 🔒 |
| Earlier urban systems work | Building-scale urban analytics, GloGCI, GloPPRUA, and agent/graph-based urban-renewal simulation | Repositories |
Building-scale urban analytics · Urban simulation · LLM/VLM evaluation and alignment · Agentic AI · Remote sensing and urban data
🌐 Academic Homepage · 🎓 Google Scholar · 📧 Email · 💬 QQ: 575860760