The problem
A driver-scoring engine was buried inside a shared service, so it couldn’t scale, deploy or be tested on its own. Three other services depended on its results, and the API needed a major version change.
What I did
Designed and led pulling the scoring engine out into a standalone Java/AWS service. It scores about 200–300k trips a day (70–80k drivers a month) for 3 services that use its results.
Load-tested at 10,000+ drivers. I found slow, incomplete DynamoDB writes and redesigned how they’re saved, batching them and processing in chunks. That cut processing time significantly and eliminated timeouts.
Built version-based routing so existing integrations kept working through a major API version change. I coordinated the mobile, DevOps and API Gateway changes, and the rollout caused zero disruption.
Worked with an external client to integrate their scoring library, with secured APIs (OAuth/JWT) and automated deployments.

Result
The service now scales, deploys and is tested independently, and every team using it moved to the new API version without an outage or breaking change.
Stack: Java, Spring Boot, Jersey (JAX-RS), AWS (DynamoDB, S3, EFS, API Gateway, CloudWatch)
0 likes · 0 views · Nov 15, 2026
Pulled a high-volume driver-scoring engine out into its own Java/AWS service, fixed its database-write bottleneck, and moved every consumer to a new API version