SUDIPTA BANIK

Designed and implemented automated fault isolation for Apache distributed job execution system using circuit breakers and adaptive retries, reducing incident Mean Time To Resolve (MTTR) to under 60 seconds and preventing cascading failures.

Built a dynamic workload admission control service, leveraging Kubernetes ConfigMaps for real-time policy propagation, successfully blocking ~7% of total concurrent job volume to protect high-priority jobs during system stress.

Improved job scheduling correctness in a distributed execution engine by centralizing ID generation and enforcing submission-time ordering, eliminating allocation delays of up to 20 seconds for ML and data workloads.

Developed a ZooKeeper-backed federation controller for the Yet Another Resource Negotiator (YARN) Router, enabling quick cluster isolation and re-attachment to maintain system Service Level Agreement (SLA) during system outages.

Enhanced Uber's internal Kubernetes batch job observability dashboard with workload-type and job-ID search capabilities and a resource-usage heatmap, improving on-call debuggability and reducing time-to-insight during incidents.

Contributed to re-designing loan offer and credit limit generation from a monthly batch to real-time event-driven processing, reducing DB load by ~30% across millions of drivers and increasing loan product adoption.

Reduced loan creation API p95 latency by almost 32% through MySQL batch inserts for installment, significantly improving throughput during high-concurrency database writes by optimizing network round-trips and lock acquisitions.

Developed a sub-second data aggregator API, integrating 5+ internal APIs for banking partners with partial response fallback, ensuring high availability and seamless data delivery.

Improved internal Kafka consumer framework reliability via graceful shutdown re-queuing, effectively reducing message loss to near zero across critical financial transaction pipelines.

Mentored 2 new engineers through comprehensive onboarding, successfully maintaining team velocity and accelerating productivity during a critical hiring season.

Designed and scaled the Refund API for Grab's Indonesia PayLater launch, handling 10% of the country's user base and gracefully managing 3M+ users through idempotent and state machine design patterns for automated retries and error recovery.

Built a tiered service fee module for real-time charge APIs and batch billing systems, implementing complex slab-rate logic to ensure compliance with stringent Indonesian regulatory requirements.

Designed a configurable credit risk assessment module in collaboration with product teams and external credit bureaus, facilitating the integration of country-specific requirements across multiple Southeast Asian markets.

Designed and implemented a robust lending credit score API, integrating data from data-science models and user metadata teams, ensuring adherence to REST API design and security best practices for financial integrity.

Improved CI/CD build speeds by ~18% and increased unit test coverage by ~35% through Go version migration, go mod adoption, and systematic refactoring, enhancing development efficiency and code quality.

Architected a Redis-backed caching layer for direct integration with 10+ global maritime carriers, significantly reducing reliance on costly third-party data providers and guiding an intern for fast-track delivery.

Resolved search disruptions by replacing a shared Elasticsearch dependency with a dedicated Port Auto-complete API, utilizing composite indexing and in-memory caching for 80,000+ ports to achieve sub-100ms latency and operational independence.

Designed an async Kafka-based maritime event enrichment pipeline, integrating data from multiple internal microservices into a unified callback payload to ensure event persistence, configurability, and reliable real-time shipment updates for customers.

Eliminated error-prone QA processes by replacing SSH-based Ruby script execution with an internal shipment event simulation tool, improving data accuracy and reducing manual effort.

Added ocean-specific features to the shipment cloning module, enabling domain-specific data replication that improved accuracy and impact of new client sales demos.