Introduction to Modern Agent Implementation (Slides)
This deck is my systematic blueprint for building modern LLM agents—in the same spirit as a course-style introduction to how LLMs are implemented, but aimed ...
About Me
I am an AI Agent Engineer at WeRide. I lead Smart Triage — an agent runtime that converts expert-driven incident diagnosis for autonomous driving into scalable, tool-grounded autonomous pipelines. The system combines dynamic skill retrieval, structured reasoning, multimodal analysis, and distributed execution so large incident volumes can be investigated with higher consistency and far lower cost than manual annotation.
I hold an M.S. in Computer Science from the University of California, San Diego and a B.S. in Computer Science and Data Science from the Hong Kong University of Science and Technology. I previously interned at WeRide and SmartTwigs.
I co-authored VisualNeo, published in PVLDB (VLDB’23), which bridges visual graph query interfaces and graph query engines so non-experts can construct and explore complex graph queries. You can access the paper here.
This webpage is last updated on Oct. 11, 2026.
Research Interests
Agents, Machine Learning, and Large Language Models — especially agentic systems for high-stakes environments, where models must use tools reliably, orchestrate complex workflows, and augment expert decision-making rather than only generate text.
Education
1. University of California, San Diego (UCSD), Sep. 2023 - Mar. 2025
- Master of Science in Computer Science (MSCS)
- CGA: 3.770/ 4.0
2. Hong Kong University of Science and Technology (HKUST), Sep. 2019 – Jun. 2023
- Bachelor of Science in Data Science and Technology, and in Computer Science
- CGA: 3.966/ 4.30
- Major CGA: 4.024/ 4.30
Awards & Prizes
- Chinese Mathematical Olympiad (CMO) — Silver Medal
- Chinese Physics Olympiad (CPhO) — First Prize
- National Olympiad in Informatics (NOI) — Second Prize
- Dean’s List for the School of Science for all grading semesters.
- Epsilon Fund Award (18th)
- HKSAR Government Scholarship
- Chern Class Award/Scholarship
- Tin Ka Ping Scholarship (Exchange)
- HKSAR Government Scholarship Fund - Reaching Out Award
- Second Prize in National Middle School Students’ Capacity Show in Maths, Physics and Chemistry (10th)
Skills
- AI Systems: Agent Runtime, Agent Harness, ReAct, Context Engineering, RAG, LangGraph, LangChain
- Cloud & Infrastructure: Docker, Kubernetes, Helm, PostgreSQL, Grafana, Prometheus, Redis
- Application Development: React, FastAPI, Next.js, Node.js, gRPC, REST APIs
- AI Models: Machine Learning, Deep Learning, Large Language Models (LLMs), Vision-Language Models (VLMs)
- Languages: Python, C++, TypeScript, Go, Java
Recent Posts
What is indeed Retrieval-Augmented Generation (RAG)?
Why I Wrote This When I first learned about RAG, I thought it was surprisingly simple. LLM + Vector Database = RAG Almost every tutorial started with embeddings, vector databases, and semantic sear...
About Modern RAG Pipelines
Why I Wrote This In my previous post, I argued that RAG is fundamentally a search system for language models. That raises a much more interesting question. If RAG is really a search system, what do...