Brett LovedayEUC & applied AI Download CV Get in touch
Selected work

Applied AI

A private AI assistant for investment due diligence.

Two investment firms needed a quicker way to answer recurring investor questions. I built an assistant that drafts answers from each firm’s own records and runs inside its infrastructure.

The question

Investment teams receive detailed questionnaires from prospective investors. A question about the firm’s operations or controls can mean searching policies, previous responses and other internal documents, then finding the right colleague to check the answer.

Both firms needed to keep their documents on their own systems. Their teams also needed to edit the draft answers in the Word or Excel files they already used.

From internal records to a first draft

I built a document pipeline that scans for new or updated files, removes duplicates and indexes the text. A question retrieves related passages through Qdrant, a database that can find passages with similar meanings even when the wording differs.

A language model running locally through Ollama uses those passages to draft a response. Staff can review the wording and export the results to Word or Excel. The files, search index and model stay in the firm’s environment.

What changed for the team

The firms adopted the assistant in their questionnaire workflows. Reviewers start with a draft drawn from internal material and can spend their time checking whether it answers the investor’s question accurately.

Discuss a project