All work

AI retrieval2025 — present

Knowledge Retriever AI

StackWise · Enterprise RAG engine

Knowledge Retriever AI — screenshot of the live site

Deep RAG

Retrieval

100s

Documents indexed

Source

Cited answers

The problem

Answers existed inside the organisation but were spread across disconnected data silos. People spent hours searching for things that had already been written down, and often rewrote them instead.

What I did

  1. 01Built a retrieval layer over the existing document estate rather than migrating it, so nothing had to move.
  2. 02Hybrid search combining lexical and semantic matching, so both exact terminology and paraphrased questions resolve.
  3. 03Every answer is returned with its source citation, making the output auditable rather than merely plausible.

Outcome

Search that people trust enough to actually use — questions resolve against the organisation’s own documents, with the reference attached.

Stack

PythonLangChainVector searchHybrid search

Senior Software Engineer · StackWise