AI-Powered Personal Photo Search Engine
2025- Built multi-stage AI pipeline indexing 46,000+ personal photos across five metadata dimensions (captions, keywords, GPS location, facial identity, date) into a unified SQLite database queryable via natural language
- Generated natural language captions for all photos using BLIP-2 on local GPU, then embedded them for semantic search
- Extracted and clustered 50,000+ face embeddings for identity grouping using DeepFace and HDBSCAN
- Built Flask web app for semi-supervised face labeling with nearest-neighbor suggestions and review workflows

