Built to move past the surface-level buzz around AI and understand the architecture firsthand. Created a document-aware chat application that lets users upload files and ask questions grounded in their own content.
- Designed the full ingestion pipeline, including upload, chunking, embedding, and semantic retrieval, using pgvector within a PostgreSQL-based stack.
- Built a cohesive product with threaded conversations, context-aware responses, and document ingestion, not just a basic chat interface.
- Deployed as a solo project, supporting document-aware chat across multiple file types including PDF, DOCX, and plain text.
