Case Studies
Every engagement here is a real project with real outcomes. Clients are anonymised where requested.
Series B Neobank · India
Built a Kafka-native fraud scoring engine with a sub-5ms p99 latency. The ML model runs on a custom Flink operator, scoring 50,000 transactions per second with no batch windows. Result: 73% reduction in chargebacks within 6 months of go-live.
Pan-India Diagnostics Chain
Deployed a FHIR-compliant NLP pipeline that structures free-text lab reports into HL7 resources. The system processes 2 million reports per month, enabling downstream analytics and seamless EHR integration across 80+ hospitals.
Fortune 500 BFSI · APAC
Lifted and shifted a 40 TB Hive warehouse to Delta Lake on Databricks without downtime. Parallelised migration using spark-copy with schema evolution handling. Query latency dropped 8× and compute costs fell 60% in the first quarter.
Enterprise SaaS · Singapore
Designed a multi-agent RAG system over 500GB of internal documentation, support tickets, and API specs. Agents can route, rerank, and synthesise across heterogeneous sources. Deployed in 8 weeks; reduced Tier-1 support load by 65%.
NBFC · Tier-2 India Expansion
Built an alternative credit scoring model using telco, UPI, and GST data for customers with no formal credit history. The feature store is powered by Feast with Postgres as the offline store. Default rates dropped 14% vs. bureau-only baseline.
Teaching Hospital · Bangalore
Streamed real-time vitals from 120 ICU beds into a Spark Structured Streaming pipeline, scoring each patient every 30 seconds against an ONNX-exported sepsis prediction model. Clinician alert dashboard in Grafana. Adverse events down 30%.
Let's Build Together
Whether you're starting from scratch or scaling an existing system — we align with your team to deliver measurable outcomes, fast.
Response within 24 hours · NDA available · No commitment required