AI · RAG2025 — Present

AI Research Insights & Fund Document Assistant

Product owner & developer · Decode Global

Two AI features built with guardrails first. The document assistant answers questions over the fund's Information Memorandum and FAQs, citing the document, section and page for every answer — and says plainly when the answer isn't in the approved documents. The User Insight CRM, built in two days for marketing's client research, turns an Excel tracker into a structured system where an LLM summary is generated on save but never overwrites the source notes.

Confidential company project. Built at Decode Global for regulated financial-services businesses — the platform and its data are private, so there's no public link or product screenshots. Happy to walk through the architecture and my contribution on request.

What it does

  • Documents parsed and chunked with fund, version, page and section metadata; retrieval filtered by fund.
  • Every answer cites document, section and page; the assistant never gives investment advice.
  • User Insight CRM: imports the team's Excel tracker, keeps raw notes append-only and links each insight to its evidence.
  • LLM summaries generated on save via structured output and stored separately, so AI output never overwrites the source.

Illustrative UI — confidential project, no real screenshots published

Next project
CeFi ↔ CRM Integration & Withdrawal Controls
→