PolyFinances Datathon 2025

RegulAI – Portfolio Intelligence

Turns dense regulations into actionable portfolio guidance by orchestrating AWS Bedrock, Comprehend, and live Yahoo Finance data inside a React/Vite experience.

← Back to projectsView GitHub ↗Live Site ↗

Architecture at a Glance

Users upload or paste a regulation inside the React front-end. API Gateway proxied Lambdas clean the text, store a raw copy in S3, then invoke Bedrock (summaries, tone, impact) and Comprehend (entity extraction). The resulting entities drive Yahoo Finance lookups, which feed a portfolio engine that scores sector/asset exposure and recommendations. Insights render in the dashboard and are mirrored to AWS QuickSight for BI.

Demo Walkthrough

After PolyFinances Datathon 2025 we no longer have AWS Bedrock access, so this recording captures the full live workflow— uploading regulations, running AI analysis, and reviewing portfolio guidance.

Prefer an interactive run-through?Visit regul-ai.vercel.app ↗

Key Highlights

  • LLM summary + tone analysis with AWS Bedrock, combined with entity extraction via AWS Comprehend.
  • Market enrichment layer pulls tickers/sectors from Yahoo Finance and normalizes them for portfolio scoring.
  • Custom risk engine ranks sector/holding exposure and renders call-to-action insights in the React dashboard.
  • Demo dashboards plus AWS QuickSight snapshots used in PolyFinances Datathon judging session.

My Contributions

  • Architected the end-to-end serverless workflow (API Gateway → Lambda → Bedrock/Comprehend).
  • Implemented the React/Vite front-end with animated insights, uploads, and QuickSight hand-off.
  • Developed the portfolio risk calculation layer (sector scoring + exposure deltas + hedging guidance).
  • Produced demo assets & recorded walkthrough once the Datathon AWS credits expired.