Client retained at $18K/month · Phase 3: AI merchant risk scoring & intelligent payment routing (in progress)
Who we worked with
A US-based payment processing company with approximately 300 employees, handling $150M+ in monthly transaction volume across 2,000+ merchant accounts. The company provides payment gateway services to mid-market e-commerce businesses.
Their core payment platform — built 11 years ago on .NET Framework 4.0 with a monolithic architecture — was reliable but increasingly painful, losing merchants to faster competitors while fraud slipped through manual reviews.
A platform losing ground to faster competitors
The monolithic architecture was creating compounding pain — slow transactions, infrequent deployments, and manual fraud review that missed 35% of fraudulent transactions.
Transaction Latency Killing Merchants
Processing averaged 2.8 seconds — competitors were under 500ms. Merchants were actively threatening to switch providers.
2.8s average latencyHigh-Risk Deployments
Releases happened once every 3–4 weeks, requiring a 2-hour maintenance window at 2 AM with the entire engineering team on call.
3–4 wk deploy cycleManual Fraud Review Falling Short
8 analysts reviewed flagged transactions one by one, catching only 65% of fraud — $2M+ in annual losses slipping through.
65% detection rateManual Scaling for Peak Loads
Black Friday required provisioning extra servers 2 weeks in advance — and the system still degraded above 800 transactions per second.
800 TPS ceilingAging PCI DSS Compliance
PCI DSS Level 1 was maintained, but audits were taking 6 weeks and growing harder as the codebase aged and documentation lagged.
6-week audit cyclesAI Ambitions Blocked by Architecture
The CTO wanted real-time AI fraud detection — but the monolith couldn't support the sub-100ms inference latency required for pre-auth scoring.
Two overlapping phases, zero downtime
A 22-week phased delivery — re-architecting the payment platform while building the AI fraud pipeline simultaneously, with transactions processing throughout the entire migration.
- Decomposed the payment monolith into 11 microservices: transaction gateway, merchant management, settlement, reconciliation, dispute resolution, reporting & more
- Migrated to AWS with auto-scaling: EKS (compute), Aurora PostgreSQL (transactional data), DynamoDB (session state), ElastiCache (hot data)
- Event-driven architecture with Kafka — enables real-time processing and independent service scaling
- New React merchant dashboard — real-time transaction monitoring, settlement tracking, self-service reporting
- Zero-downtime blue-green deployments via GitHub Actions — from 2 AM maintenance windows to multiple daily releases
- End-to-end encryption, tokenization, comprehensive audit logging for PCI DSS Level 1 compliance
- Real-time fraud scoring pipeline — every transaction scored by ML model within 45ms of submission, before authorization
- XGBoost gradient boosting model trained on 3 years of labeled transaction data, using 47 features: amount, device fingerprint, geolocation, velocity, merchant category & behavioral patterns
- Adaptive learning pipeline — model retrains weekly on new labeled data (confirmed fraud + false positives) for continuous improvement
- Fraud analytics dashboard for risk team: real-time fraud rate monitoring, geographic heat maps, merchant-level risk profiles
- Merchant anomaly detection — flags accounts with sudden transaction pattern changes indicating potential compromise
47 engineered features
Legitimate vs. fraudulent
New fraud patterns incorporated
Built for speed, scale & compliance
Results that transformed the business
Across performance, fraud prevention, compliance, and team efficiency — all delivered in 22 weeks with zero transaction downtime.
Transactions
Detection Rate
Losses Prevented
Rate (was 12%)
7 engineers, one cohesive delivery
A specialized team combining payments domain expertise with modern cloud engineering and ML — structured for complete ownership across the full stack.
Solution Architect
Payments domain expertise, microservices design, AWS architecture
Senior .NET Developer × 2
.NET 8 microservices, payment gateway logic, API design
Senior React Developer
Merchant dashboard, real-time monitoring UI, reporting interfaces
ML Engineer
XGBoost fraud model, feature engineering, Kafka Streams scoring pipeline
DevOps / Infrastructure Engineer
Kubernetes, Terraform, blue-green deploys, PCI DSS compliance logging
QA Engineer
End-to-end test automation, performance testing, payment flow validation