AI Product Consultant

01 — Business Problem

Customers struggled to choose between similar products, and a plain LLM invented specifications.

02 — My Role

I designed the retrieval pipeline, prompt strategy and API layer, and tuned answer quality on real queries.

03 — Solution

Catalog and knowledge base content is chunked, embedded and retrieved before generation, so answers cite real attributes.

04 — Architecture / How It Works

Ingestion job · vector store · retrieval service · LLM generation · REST API for the frontend.

06 — Challenges

Keeping answers grounded while data changes daily, and controlling latency and token cost.

07 — Result

Faster product selection with markedly fewer invented specifications.

05 — Technologies