← Applied AI engineering

Internal product

Seller RAG

Product knowledge with a deliberate access boundary

Applied AI architecture and implementation
Recent confidential engagement

  • Python
  • C#/.NET
  • Semantic Kernel
  • Cosmos DB
  • RAG

The problem

The client needed product and operational information in one place, but did not want an AI system making consequential changes to the Seller account.

What I did

  • Use a read only Claude Code Skill to orchestrate Python retrieval.
  • Build the knowledge path in a separate C# Semantic Kernel application.
  • Store product information and embeddings in Cosmos DB for NoSQL with the serverless connector; retain manual authority over changes.

The result

Internal users trialled the product knowledge application.

Scope and boundaries

The client and account remain anonymous. This is not a claim of autonomous Seller updates, sustained production operation or measured business impact.

Portfolio summary updated September 2026.

Talk to me about similar work