Case 01 — Quantzig × Shell
Enterprise Knowledge Copilot
- Year
- 2026
- Role
- Full-stack engineer — retrieval to interface
17A
A conversational interface that turns a sprawling enterprise knowledge base into answers you can trace back to a source.
- Conversational AI
- RAG
- Next.js
- Python
Problem
Shell's internal knowledge lived across thousands of documents. Finding the right answer meant knowing where to look, and trusting it meant reading the whole thing.
Approach
- 01Designed a RAG retrieval pipeline that chunks, embeds and ranks documents so answers stay grounded in sources.
- 02Orchestrated LLM calls behind a typed API layer, with streaming responses and citation metadata.
- 03Built the chat interface so every answer shows where it came from, which made people trust the tool enough to use it.
Outcome
Natural-language search over the knowledge base, with answers cited back to their source documents.
Some of this work is under NDA, so these notes describe the problem, the craft and my contribution without private product details.