Mastercard
Payments infrastructure · via Endava
Fintech2023
Senior Software Engineer · via Endava
Context
Mastercard’s multitenant payment infrastructure: many internal consumers hitting the same APIs, with latency and tenant isolation requirements that are not negotiable.
The challenge
Process financial events in near real time while giving internal teams reusable APIs and per-tenant analytics, without any single consumer being able to degrade the others.
What it does
- Multitenant API gateway with rate limiting
- Near real-time financial event ingestion
- Fraud detection and risk scoring
- Per-tenant analytics on OLAP
- Internal transaction monitoring dashboards
What I did
- Designed internal APIs and reusable backend components: Python API gateways and middleware for routing, rate limiting and tenant-scoped authentication.
- Built Python consumers and producers over Kafka and gRPC, with Protobuf schema validation, for the low-latency event ingestion system.
- For fraud and risk scoring we built an OAuth-secured API that only the model consumed: it got the OpenAPI spec in context plus a Neo4j graph with the base data, so it could interpret the relationships, reach its own conclusions and decide which endpoints to call. Credentials and endpoints came from environment variables per tenant and environment, so access was scoped and keys rotated without touching code.
- Designed schema strategies for ClickHouse OLAP tables, optimizing cross-tenant analytics workloads.
- Built and maintained ingestion and transformation pipelines with Apache Airflow, authoring the DAGs for ETL flows and reporting layers.
- Developed internal React tools for transaction monitoring: visualizing payment flows, spotting anomalies and drilling into per-tenant metrics.
Stack
- Python
- API gatewayMiddlewareKafka consumersApache Airflow
- Messaging
- Apache KafkagRPCProtobuf
- Data
- ClickHousePostgreSQLETL
- JavaScript
- ReactDashboards
- AI
- Neo4jOpenAPITool useRisk scoring
- Multitenant
- Rate limitingTenant authData isolation
Services practised here
- Platforms & Distributed SystemsWhen a product grows, the problem stops being a feature and becomes how the services talk to each other. That is where service boundaries, event delivery and flows that cannot be lost along the way come in.
- AI Integration & AutomationModels and agents go into the product you already have with the same discipline as any other service: a contract of your own, cost limits and a log of what they did.
Outcome
Three years on the same platform. Multitenant design stopped being a one-off architectural decision: every endpoint, table and dashboard is designed knowing which tenant each piece of data belongs to.