Staff Software Engineer — payments, fraud and risk systems Go · Java · Python · Kafka · Distributed systems · Applied ML

I build the systems that move money — and lately, the ML that decides whether the money should move at all.

Eleven years across the length of a bank: KYC and customer onboarding, real-time payment initiation and settlement, ACH and card scheme processing, credit risk and loan origination. Currently at Q2, building AI-driven fraud and anomaly detection across ACH payment, batch and collection flows — catching suspicious activity before settlement, not in next-day reconciliation.

Before that: Confluent (led a real-time intelligence platform from zero to MVP in five months), Cisco, DBS Bank, American Express transaction processing, and a Swiss retail bank.

Open to Staff / Senior Backend roles. Currently exploring opportunities in the UK.

📩 sourav.mansingh5@gmail.com 🔗 LinkedIn 🌐 souravmansingh.com ✍️ Medium

The full transaction lifecycle — initiation, authorisation, capture, refund

A distributed payment platform built the way a real one has to be: idempotency keys so a retried request never double-charges, saga orchestration so a half-completed payment can always be unwound, and Kafka for the async legs that must not block the customer.

The hard part isn't throughput — it's being wrong safely.

Java · Spring Boot · Kafka · PostgreSQL

Scoring a transaction's risk while it's still in flight

Tiered fraud scoring over a streaming transaction feed — a fast gradient-boosted model for the common case, escalating to sequence and graph models where the signal warrants it. Adverse-media sentiment from SentimentPulse feeds in as an optional feature.

Python · Kafka · XGBoost · FastAPI

Real-time financial news sentiment, end to end

Ingestion through to a live dashboard: Kafka pipeline, a FinBERT model fine-tuned on financial text (0.92 F1, published on Hugging Face), TimescaleDB for the time series, FastAPI and React on top.

Built during a deliberate break from full-time work, to learn what it actually takes to put a model in a pipeline rather than a notebook.

Python · Kafka · FinBERT · TimescaleDB · React

Messaging, presence and WebRTC signalling

Goroutine-per-connection, Redis pub/sub for cross-node fanout, Postgres for durable history. Written to get properly fluent in Go's concurrency model.

Go · WebSocket · Redis · PostgreSQL

I write about distributed systems, payments architecture and applied ML on Medium.