I built SkillVector, a live AI career intelligence platform with a three-layer retrieval pipeline. By adding an in-memory cache, pgvector search, and web fetching, I reduced warm-query latency from 9.6 seconds to 0.48 seconds.
I also built Flash Resume, a local-first resume tailoring tool that reached 850 PyPI downloads and about 20 active users. Through the Wikimedia Student Engagement Program, I built a community insights dashboard and reduced SQL query execution time by about 25% on replicated Wikimedia datasets.

