FintechData & ML platform
Task. A payment service in Tashkent was growing fast: transaction data sat in different systems, antifraud rules fired too late, and regulator reporting was prepared manually.
Solution
We built a Data & ML platform: streaming transaction collection, antifraud and scoring marts, real-time anomaly-detection ML models, automated reporting. Deployed inside the client’s perimeter in Uzbekistan.
Results
2×faster antifraud decisions
5×less effort for report preparation
1.5×more accurate scoring criteria
Technologies: Kafka, ClickHouse, Spark, MLflow, Python, Kubernetes.
Project stages
- Architecture (2–3 weeks) — data-platform blueprint for the payment service.
- Core build (6–10 weeks) — streaming pipelines, warehouse and ML infrastructure.
- Use cases (4–8 weeks) — fraud detection, scoring and personalisation models.
- Operations — MLOps, monitoring and team enablement.
Where this applies
Data/ML platforms power fintech in Uzbekistan and Kazakhstan, payment services in Kyrgyzstan and e-wallets across Central Asia. See data management & analytics.