Work / RAG & knowledge / Live

Compliance knowledge assistant for a fintech team

An internal assistant that answers policy and regulatory questions from the company's own documents, with citations and access control.

Problem

Operations and support staff spent time searching long policy documents and circulars, and answers varied depending on who was asked.

Objective

Give staff fast, consistent answers grounded in approved documents, always showing the source, and never exposing documents a user isn't allowed to see.

Solution

  1. 01

    Documents are parsed, split by section and indexed with both embeddings and keyword search.

  2. 02

    Retrieval filters by the user's permissions before anything reaches the model.

  3. 03

    Every answer cites the document and section it came from; when nothing relevant is found, the assistant says so.

  4. 04

    An admin view lets the compliance team add, replace and retire documents.

Architecture

01
Staff web app (SSO)
02
API
03
Permission filter
04
Hybrid search (pgvector + keyword)
05
LLM
06
Document store

AI workflow

01

User asks a question

02

Search only documents the user can access

03

Rank and select the most relevant passages

04

Generate an answer that cites each source

05

Log the question and sources for review

Technology & role

Python / FastAPIPostgreSQL + pgvectorClaude APINext.jsSSO
Our role

Retrieval design, evaluation set, backend, admin tools and deployment.

Start here

Want something similar built?

Send a short brief. We'll reply personally with questions or a suggested next step — usually a technical discovery call.