Arijit Mitra
← Home

Fintech Systems

Setu Payment Service

Production-minded payment lifecycle event ingestion, immutable ledger state machine, and SQL reconciliation engine.

FastAPIPythonPostgreSQLSQLAlchemyFintechReconciliation

The problem

Asynchronous webhook hops, network retry storms risking catastrophic double-credits, fragmented multi-day transaction states, and financial reconciliation discrepancies between gateway payments and actual settlements.

Approach

  • Engineered an event-driven immutable ledger using FastAPI and Supabase (PostgreSQL via SQLAlchemy) deployed serverless on Vercel.
  • Implemented strict database-enforced idempotency on event_id — safely swallowing network retries with a 200 OK already_processed response to prevent double-credits.
  • Built a state machine engine dynamically deriving the master transaction state across payment_status, settlement_status, and lifecycle progression.
  • Pushed analytical calculations directly down to PostgreSQL using optimized GROUP BY and CASE WHEN queries to aggregate merchant settlement rates and transaction volumes in milliseconds.
  • Developed a dedicated anomaly detection engine classifying real-world financial edge cases: settled_after_failure (potential fraud), processed_not_settled (stuck money), and stale_initiated.
  • Shipped a branded Developer Experience (DX) portal with ReDoc, interactive Swagger OpenAPI, visual web dashboard, and a 24-point comprehensive Postman test suite.

Impact

  • Zero double-credits guaranteed via database-level idempotency
  • Real-time SQL discrepancy detection across 3,800+ transactions & 5 merchants
  • Complete live developer experience: Base API, ReDoc guide, Swagger & visual dashboard
Next projectCredit Interceptor →
Setu Payment Service | Arijit Mitra