PSB-Hackathon-Series-2026 — SecureWealth Twin
SecureWealth Twin — Intelligent Wealth Growth with Built‑in Fraud Protection
An end‑to‑end prototype combining personalized AI wealth advisory with a mandatory fraud‑protection layer for critical actions. This README guides contributors and judges through setup, architecture, components, development workflow, testing, and deployment.
Table of contents
Project overview Key features Tech stack Repo structure Prerequisites Step 1 — Backend setup (detailed) Step 2 — Frontend setup (detailed) Environment configuration (.env) Database schema (summary) Running locally (both services) API docs & Postman collection Tests (unit, integration, load) Mock/demo data Security & secrets CI/CD (Cloud Build sample) Deployment notes (GCP Terraform pointers) Troubleshooting Contact / Contributors Project overview SecureWealth Twin is a demo-grade platform showing how banks can offer AI-driven personalized financial advice while enforcing mandatory fraud-protection for transfer, trade, withdrawal, and payout setup flows. The repo contains a FastAPI backend, React frontend, and demo assets for the PSB Suraksha Manthan hackathon.
Key features (high level)
Onboarding with KYC stub, risk profiling, and simulated account tokenization Personalized wealth advisor: goals, allocations, rebalancing suggestions, scenario simulation, natural language Q&A (LLM) Fraud Protection (mandatory): real‑time risk scoring, device fingerprint basics, velocity checks, step‑up auth, human approval / cooling period, transaction throttling Explainability & Audit: immutable audit entries, reason codes, customer-facing risk messages Monitoring hooks, SIEM-friendly logs, demo incident runbook https://ai-wealth-shield.preview.emergentagent.com/dashboard Demo mode with synthetic users for hackathon presentation
Tech stack
Backend: Python, FastAPI, Uvicorn, SQLAlchemy, Alembic, Celery (optional), Redis (optional) Auth: Firebase Authentication (email/phone/SSO) — simulated locally / stubbed for demo Database: MySQL (production), SQLite for quick local dev Frontend: React (web) and React Native (mobile) skeleton (mobile not included in this repo—see mobile/ folder) Serverless/Cloud: Cloud Functions (business logic), Cloud Run (AI microservice), Pub/Sub (events), BigQuery (analytics) — infra via Terraform (samples) LLM: microservice pattern calling instruction‑tuned LLM (OpenAI or other) with embeddings for profile + history Observability: Stackdriver / Cloud Logging (GCP), optional SIEM integration Secrets: Google Secret Manager (production); local uses .env (never commit) CI/CD: Cloud Build (sample cloudbuild.yaml) + GitHub Actions.