I'm fascinated by the gap between a great idea and a system that actually holds up at scale. That gap is where I like to work.
I recently finished my MS in Computer Science at San José State University, where my research on image-based malware classification became a co-authored paper, now under review. Before grad school I was a software engineer on an EdTech platform serving 70K+ users, chasing down latency and shipping features thousands of learners used every day.
Outside of work I take systems apart to understand them. Lately that meant writing a vector search engine from scratch in C++, then building an LLM gateway and a movie recommender on top of it. I ship something small almost every day; building is how I learn.
Open to full-time Software, ML, and AI Engineering roles across the US.
| Project | The short version | Code |
|---|---|---|
| Proxima | A C++ vector search engine that answers queries 1.8× faster than hnswlib and 2.5× faster than FAISS at 0.999 recall | repo |
| Relay | An LLM gateway that remembers: 78% of requests served from a semantic cache, median latency 759 ms to 44 ms | repo |
| Reel Rank | A two-stage hybrid movie recommender with retrieval running on Proxima, answering free-text requests like "a slow-burn sci-fi like Arrival but funnier" | repo |
| Autograde AI | A local-first multi-agent grading platform: six grader agents under Temporal and Kafka, confidence-gated human review | repo |
| Malware Classification | My thesis: malware binaries rendered as images, a three-track ensemble that agrees 94% of the time across 17 families | repo |
The full case studies, with architecture and honest numbers, live on the portfolio.