Prabhath akula

Migrated legacy background job systems to Temporal workflows, improving observability by 60%, enhancing system reliability, and increasing scalability by 3x across distributed services. Built a self-serve platform for onboarding lenders into the Yubi CoLending platform, cutting UAT initiation time by 50-70% per new partnership and eliminating manual SDLC bottlenecks. Developed and deployed high-throughput Kafka consumers, enabling seamless real-time processing of over 10 million messages/day, improving latency and system throughput. Implemented secure SFTP pull/push automation, integrating with over 15+ client SFTPs to enable encrypted data exchanges, improving file transfer reliability by 40%. Designed and deployed custom rate limiters and cron-based schedulers, meeting diverse investor SLAs and reducing task execution load by 30%, while ensuring high system uptime. Led GRPC API orchestration efforts for a top-tier bank, eliminating customer duplication issues across services using GRPC-based communication, improving customer data accuracy by 95%. Drove a major observability upgrade by streaming distributed traces into Apache Pinot for real-time insights. Technologies used: AWS (EC2, S3, RDS), Temporal, New Relic, Kloudfuse, Kafka, Zookeeper, MySQL, Kubernetes, Docker, Redis, GRPC, SFTP, CRON.

Co-developed a self-service ML platform enabling internal teams to upload datasets, train models, and deploy pipelines with zero infrastructure setup, empowering 50+ non-technical users to leverage AI tools independently. Integrated Amazon SageMaker with Ray to facilitate distributed training on datasets exceeding 100GB, reducing average training time by 65% and enhancing scalability for complex ML workloads. Stack used: Python, Django, TensorFlow, Scikit-learn, Pandas, Amazon SageMaker, Ray on AWS.