PolicyGuard.Ai
Link to open source: https://github.com/Mrvikas06/policyguard-aml-dashboard
Link to Live Project: https://mrvikas06.github.io/policyguard-aml-dashboard/
🔷 Problem Statement
"Financial crime doesn't leave notes. It leaves patterns."
"Manual AML review is like finding a needle in half a million haystacks — daily."
"Every $1 of money laundering cleaned costs $5 to investigate. We flip that equation."
🔷 Solution
"Upload a policy PDF. Watch it become 25 enforceable rules in seconds."
"From document to detection — in one pipeline."
"We don't just flag violations. We explain them, score them, and file the paperwork."
🔷 IBM AML Dataset
"487,312 real-world transaction patterns. One ground truth column. Infinite edge cases."
"Trained on IBM's synthetic AML data — the same complexity as real financial crime."
"Every violation we detect is backed by is_laundering = 1."
🔷 Technology
"LLMs read policies. SQL catches patterns. Humans make the call."
"50ms from transaction to alert. Faster than a compliance officer can open their inbox."
"Claude reads the rules. The database enforces them. You stay in control."
🔷 Results / Demo
"80% recall. 80% precision. 10 violations. 0 missed SARs."
"We don't replace compliance officers — we give them superpowers."
"From PDF upload to FinCEN SAR filing — in one dashboard."
🔷 Closing / Vision
"Every bank has a compliance team. We give them an AI co-pilot."
"PolicyGuard.AI — because the next money laundering network is already running."
"The future of compliance isn't more analysts. It's smarter systems."
This build was uploaded as a hackathon project




