Context and problem

At an electric-charging equipment manufacturer, product and technical documentation grew quickly along with the charging infrastructure. Specialists needed reliable access to the manuals and specifications without every search requiring knowledge of where the information lives. They needed an internal tool that understands the content of the documents semantically, not just as keyword search.

What was done

I acted as lead developer and was responsible for the architecture and implementation of the solution. I built an end-to-end RAG pipeline: indexing from several document sources, semantic search, and context-aware answer generation. The implementation used the Amazon Q environment and RAG models for production use.

I was responsible for the NLP side (retrieval quality, scoping the context, the usefulness of the answers) and for bringing manuals of different formats together into one searchable whole. Delivery April–September 2024.

Key technologies: RAG, Amazon Q, semantic search, NLP.

Outcome

The system was taken into internal use and sped up access to product information. The technical documentation was searchable from one place without manual review, which reduced friction especially in support and development work.

← Back to assignments