At Instacart, I built backend search services that applied retailer and availability constraints to product retrieval. I tuned queries and filtering strategies to balance search coverage with database load and response latency.
I extended keyword retrieval with embeddings and pgvector-based semantic search, and integrated LLM query-understanding outputs with validation, caching, and deterministic fallbacks. I also built workflows to evaluate and monitor retrieval quality and model regressions.
At Oscar Health, I optimized Go care-search services with concurrent dependency calls and batched provider lookups, reducing response latency from approximately 500 ms to 200 ms. I also added deadlines, fallbacks, and data-freshness checks to help keep search responsive and provider information reliable.
At Pinterest, I developed Home Feed features and contributed to Kafka and Amazon S3 event-data pipelines. I helped migrate web experiences to React and server-side rendering, and improved data fetching with batching, concurrency, caching, and fallbacks.
