Benchmarked FP16, INT8, and INT4 inference across 34 configurations, measuring throughput, memory, and energy per token.

Built a Go storage engine with an LSM tree, quorum replication, consistent hashing, and failure recovery.

Built a research assistant with LangGraph, Milvus, and Gemini, deployed on Google Kubernetes Engine.

Implemented a memory allocator, a Unix shell with job control, and a concurrent file system.

Built real-time video rooms with chat, a shared whiteboard, and screen sharing using WebRTC and Socket.IO.

Explore more of my work, experiments, and source code on GitHub.