The Challenge
The bank’s fraud prevention team operated a 12-year-old rule engine with 240+ static rules manually curated by analysts. As digital transaction volumes tripled post-COVID, the system collapsed under load — scoring delays of 3-8 seconds were killing the mobile banking UX. Worse, adaptive fraudsters learned to circumvent the static rules within days of each update, causing recurring losses across credit card skimming, UPI reversal fraud, and identity-spoofed account takeovers.
Our Solution
HNBC designed and deployed a three-layer real-time fraud intelligence platform built around ensemble machine learning models, streaming event processing, and a closed-loop feedback system that continuously retrains on new fraud patterns
Architecture Highlights
The platform runs on AWS with EKS for the model serving layer, Kinesis for event ingestion, and Aurora PostgreSQL for the feature store. Auto-scaling handles transaction spikes during salary credit days (5-10× normal volume) with zero SLA degradation. All model decisions are logged to an immutable audit ledger compliant with RBI’s fraud monitoring framework.
Case Study 02 · Healthcare
Hospital Information System Modernisation
500-Bed Multi-Specialty Hospital • Chennai, Tamil Nadu
Zero-Downtime Migration from Legacy HIS to Cloud-Native Platform
A leading multi-specialty hospital was running a 12-year-old HIS that produced ₹40L+ in annual billing errors and required 45 minutes to generate daily MIS reports.
The Challenge
The hospital operated across four buildings with 47 departments, 500 beds, and 1,200+ daily outpatient visits. Their legacy HIS ran on an on-premise Windows server with a monolithic .NET codebase last updated in 2012. Critical issues included: billing module miscalculations costing ₹40L+ per year, zero mobile access for doctors during ward rounds, no interoperability with diagnostic lab systems, and a complete inability to participate in India’s Ayushman Bharat Digital Mission (ABDM) due to lack of FHIR support.
Our Solution
HNBC designed a phased, zero-downtime migration strategy: run the new cloud-native system in parallel for 60 days before cutting over. The new platform was built as a microservices architecture on Azure, with each hospital department as an independently deployable domain service.
Architecture Highlights
Data migration of 3M+ patient records was executed using a custom ETL pipeline with deduplication, validation, and rollback capabilities. The 60-day parallel run allowed staff to gain confidence before the final cutover, which happened on a Sunday at 2 AM with 100% data integrity. Post-go-live, HNBC provided on-site support for 30 days, training 380 staff members across all shifts.