Freelance data scientist · All case studies

Search that knows the session, not just the track.

A retrieval-augmented system over a listening-session database: 26 relational tables cleaned and mapped, then semantic search, voice-shaped queries, and TF-IDF as the honest baseline next to embeddings.

Client: Audio session search. Built by Dilshad Raza.

Twenty-six related extracts described users, sessions, tracks, and events. A keyword box could not reconstruct “what I played after the jazz set.” The model had to retrieve from a mapped listening graph, including queries that arrive as speech.

I parsed and cleaned the relational set, mapped listening tracks into a searchable graph, and built a RAG layer with semantic search plus TF-IDF vectors. Voice-oriented formatting sat in front so a spoken question still hit the right session rows.

Hire Dilshad Raza for similar freelance data science, machine learning, and AI automation in the UK, United States, and Australia.