Building production GenAI systems.
Previously led GenAI at MaiAgent (scaled users 3K → 20K) · LlamaIndex contributor (14 merged PRs) · M.C.S. @ Rice · Incoming SWE Intern @ Google
Former GenAI Team Lead who shipped Agentic AI Systems, Advanced RAG, and the backend infrastructure behind a B2B platform serving CTBC Bank, MSI, HPE, and iGroup — work that contributed to the Presidential Hackathon Winner (2024) and 23rd Golden Peak Award (2025). I care about systems that survive contact with real users: retrieval that scales past 20M+ chunks, agents that don't burn tokens, and test coverage that catches regressions before they ship.
Now at Rice (GPA 4.00) and joining Google as a SWE intern.
|
LangGraph · Neo4j · Milvus · Multi-Agent
Self-routing agentic system that picks Graph RAG, Vector RAG, or a hybrid path per query. Achieved 88.2% MAR@10 on multi-hop datasets and cut indexing token cost by 35% via ML-clustered pivotal node selection. |
CrewAI · LangChain · Streamlit · Chroma
Multi-agent tutoring system with role-specialized LLMs for problem-solving, exam detection, and concept analysis. Top 3 of 50 teams at the NYCU CS AI workshop (Outstanding Award). |
|
Open Source · 15K+ ⭐
14 merged PRs across AWS Bedrock, Claude, Elasticsearch, Cohere, OpenAI, MCP Client, and the AI Agent Workflow modules — bug fixes, new integrations, and capability extensions. |
Django · FastAPI · LlamaIndex · AWS
Led GenAI roadmap for a 10-person startup. Upgraded chatbots into AI agents with memory + tools + MCP, driving 120% partner growth and 67%+ token reduction. Introduced pytest framework reaching 67% coverage from zero. |
Languages Python · TypeScript · C/C++
Backend Django · FastAPI · Express · Celery · Nginx
GenAI LlamaIndex · LangChain · LangGraph · MCP
Data PostgreSQL · Neo4j · Milvus · Elasticsearch · MongoDB · Chroma
ML / DL PyTorch · JAX · scikit-learn · NumPy · Pandas
Cloud AWS (Bedrock · SageMaker · EMR · EC2 · S3 · RDS · ElastiCache)
DevOps Docker · GitHub Actions · Linux · pytest




