02 / Generative AI · RAG

Retrieval systems & internal AI assistants

Retrieval-augmented assistants over your own documents and data — built for answer quality and traceability, not just a vector database and a prompt.

Who this is for

Organisations with a real internal knowledge problem — policy documents, technical manuals, support history, contracts — where staff currently search manually or ask around, and wrong answers are costly.

The problem

A vector database and a system prompt gets you a demo, not a reliable assistant. The hard part is retrieval quality: chunking that respects document structure, ranking that surfaces the right passage, and answers that cite their source so a human can verify them.

What we deliver
  • A retrieval pipeline tuned to your document structure, not a generic chunker
  • An assistant interface (chat or API) with inline source citations for every answer
  • An evaluation set of real questions with known-correct answers, used to measure retrieval quality before and after tuning
  • A process for keeping the index current as documents change
Stack
pgvector / OpenSearchClaude / GPT-class modelsHybrid lexical + semantic retrievalAWS
Timeline

5–6 weeks

Price
From $20,000
Founding client rate $14,000
In practice
91%
of internal policy questions answered correctly against a held-out eval set
Contact

Have a workflow worth building right?

Start a project →