Inspiration Every fintech today learns about new regulations by accident — usually after a fine — and we wanted to fix that.

What it does MERIDIAN ingests live regulator feeds, maps every new rule to your internal policies, and ties it to a verifiable audit trail of every automated decision your platform makes.

How we built it React + Vite on the front, Lovable Cloud (Postgres + Edge Functions) on the back, and the Lovable AI Gateway (Gemini 2.5 + GPT-5) for policy mapping, PDF extraction, and training generation.

Challenges we ran into Wiring a multi-agent orchestration pipeline that could reliably extract policy PDFs, map them to live regulator items, and emit a strict JSON contract without hallucinating was much harder than any single piece suggested.

Accomplishments that we're proud of We shipped a real end-to-end RegTech loop — live EBA feed → AI policy mapping → auditable decision trace → exportable audit pack — in a single hackathon sprint.

What we learned Strict tool-calling schemas, security-definer RLS, and treating AI outputs as untrusted JSON are what separate a demo from something a CTO would actually plug into production.

What's next for Meridian Adding more live regulator connectors (ACPR, FCA, OFAC), per-tenant fine-tuned policy mapping, and a Slack/Jira "notify owner" loop so every new rule becomes an assignable, trackable action.

Built With

  • supabase
  • vite
Share this project:

Updates