At OCI Data Science, I designed and implemented a multi-tenant API architecture that serves multiple inference endpoints from one backend service instance, reducing infrastructure duplication by 75–80%. I also extended a stateless HTTP backend with bidirectional WebSocket support for streaming and low-latency workloads.
On OCI's Model Deployment Team, I drafted and implemented a backend for deploying and serving large ML models, reducing deployment setup time by 60%. I productized inference runtimes in a managed deployment workflow, cutting customer setup effort by more than 90%.
I’ve also improved service stability through root-cause analysis, timeouts, retries, and failure handling, raising fulfillment KPIs by 30%. I enjoy turning loosely defined engineering problems into production-ready systems, with a focus on reliability and maintainability.
