I built FilingIQ, a table-aware RAG system for analyzing SEC Form 10-K filings from Apple and Tesla. I benchmarked dense, BM25, and hybrid retrieval across 480 data chunks and 12 verified questions, achieving 0.83 recall@1.
I also built and deployed a containerized RAG service using FastAPI and Qdrant, with document ingestion and monitoring; retrieval latency was 74.79 ms. I established latency and accuracy benchmarks to evaluate its performance.
For an AI-based autonomous drone landing system, I built a pipeline that identifies safe landing sites from a single aerial photo using U-Net semantic segmentation, terrain scoring, and hard safety constraints. I’m pursuing an M.E. in Communication Engineering and have internship experience in embedded systems.
