I'm an INFJ-T passionate about technology and innovation, with a Columbia University M.S. in Computer Science and a University of Toronto B.S. in Computer Science and Statistics.
My interests center on AI and large language models, including RAG, long-term memory, model integration, and intelligent agent applications.
I'm dedicated to building practical AI systems that connect rigorous machine learning ideas with reliable software engineering and real user needs.
Columbia University, New York
2026 - 2027
University of Toronto, Canada
2022 - 2026
Built an AI NPC dialogue plugin for Minecraft servers, integrating LLM APIs, embedding APIs, RAG knowledge retrieval, and model-driven long-term memory.
Designed a self-adjusting economy plugin using player transaction logs, online value discovery, dynamic AMM pricing, and statistical learning.
Designed and implemented a hybrid provenance framework integrating TrustMark imperceptible watermarking with C2PA content credentials.
Introduced LLM collaborative development (Vibe Coding) through prompt engineering. Developed a full-stack architecture using React.js, Node.js, Express.js, Prisma, and PostgreSQL.
Created a rendering system with ray tracing, shader pipelines, and physical animation. Used dynamic BVH acceleration for efficient intersection tests.
Engineered a user-level thread library in C with cooperative/preemptive multithreading. Built a virtual memory simulator with hierarchical page tables.
Developed a comprehensive web platform for UofT's Chinese student organization, featuring bilingual support and cultural event showcases.
Led formative research through interviews and surveys. Designed and tested prototypes, increasing task success rate by 32%.
Led basketball analytics project in R. Built multiple linear regression models with rigorous diagnostics including VIF and Box-Cox transformation.
Implemented core Tetris mechanics in MIPS assembly, including block generation, movement, rotation, and collision detection.
Implemented "Clean Architecture" and advanced design patterns. Achieved 100% test coverage with JUnit 5 and integrated Todoist API.