Exploring AI research and building practical systems.
- Currently at Alibaba Taotian, based in Hangzhou, China.
- Previously at Tencent Hunyuan. A graduate of Nanjing University.
- Interested in Agentic AI, LLM Systems, and Multimodal Learning.
| Period | Experience |
|---|---|
| 2026–present | Alibaba Taotian · 阿里淘天 |
| 2025–2026 | Tencent Hunyuan · 腾讯混元 |
- Research: visual question answering, multimodal understanding, and efficient visual representations.
- Engineering: LLM inference, agent workflows, retrieval, and evaluation.
- Learning in public: connecting paper ideas with implementation details and practical trade-offs.
千面 LLM · A Thousand Faces of LLMs
Question-driven technical notes in Chinese, covering research ideas, system design, and engineering trade-offs. Start with:
- DeepSearch: query selection and challenging question construction
- Mixed RL data integration and multi-teacher online distillation
欢迎通过 Issues 交流问题、勘误和选题。
- CLI-Anything #498: align CLI harness package license metadata with Apache-2.0.
- My CLI-Anything, PRIME, and rllm forks are for exploration and development; original projects are credited in each repository.
| Area | Tools |
|---|---|
| Languages | Python · Go · C++ · TypeScript |
| Models and agents | PyTorch · vLLM · LangChain |
| Infrastructure | Docker · Kubernetes · Redis |