Freelance data scientist · All case studies

Ask the PDF. Get the page back.

A dual-mode retrieval engine: Hugging Face transformers, Chroma, MiniLM embeddings, and Microsoft Phi-2 — so an expert question maps to the passage and the page it came from.

Client: Document Q&A system. Built by Dilshad Raza.

Expert documents are long, versioned, and hostile to chatbots. The requirement was dual-mode retrieval that could return the context string and the source page — not a fluent paragraph you cannot defend in review.

I built the engine on LangChain with Hugging Face transformers, a Chroma vector store, SentenceTransformers MiniLM-L6-v2 for embeddings, and Microsoft Phi-2 for generation — always mapping the context window back to source pages.

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