MSc Computer Science student at Georgia Tech, previously CS at UBC.
I enjoy building reliable software systems and turning research ideas into working prototypes. My work spans full-stack engineering, applied machine learning, LLM systems, and AI agents.
- ML, LLM, DRL, world models, and embodied AI
- Full-stack engineering for useful AI systems
- Research projects that connect modeling, evaluation, and real workflows
- Snowboarding and teaching as a CASI snowboard instructor
- Inductify - AI-powered onboarding assistant with LangChain, RAG, ChromaDB, FastAPI, OpenAI, and Docker.
- Data Processing Web System - Fault-tolerant async processing system with Celery, Redis, PostgreSQL, SQLAlchemy, Alembic, and uv.
- UniQuery - Full-stack web application for querying UBC course data.
- On-Policy Distillation for Vision-Language Reasoning - Transferred on-policy distillation to Qwen3-VL on ChartQA: a 2B student trained on 300 questions matches supervised distillation on 3,000, with a token-level analysis of where the teacher's feedback lands. Code
- Soccer-Twos Multi-Agent RL - PPO self-play curriculum for 2v2 Soccer-Twos with reward shaping and mixed opponents.
- GNNs for Product Recommendation - Graph neural network study for Amazon co-purchase recommendation. Preprint