Site Reliability Engineer
Current* Identified backend allocation inefficiencies through async-profiler analysis and Java heap analysis, leading to a reduction in P50 latency from 20ms to 15ms.* Managed Aerospike infrastructure of 30 clusters across 4 regions handling loads upto 500K QPS, driving upgrades and performance improvements.* Planned and optimized capacity for Redis clusters, including surge management to ensure high availability during peak loads. Automated bin-packing of Redis shards across nodes, resulting in a 20% improvement in Redis node utilization and enhanced scalability.* Implemented eBPF-based CNI (Cilium) to enable cross-team cluster mesh, reducing Global Load Balancer (GLB) costs by 30%. Stabilized P99 latency for services, achieving a 40% improvement in performance and reliability* Reduced pod startup time by leveraging Go concurrency to quickly load models, resulting in a 20% decrease in average latency. Optimized garbage collection (GC) times through profiling, driving improved system performance. Conducted experiments with CPU pinning, enabling efficient application runs on machines with fewer cores, enhancing resource utilization and scalability.* Integrated StackStorm, an Automated incident response system, reducing Mean Time to Resolution (MTTR) for common alerts and issues from 10-15 minutes to just 2 minutes, significantly improving incident handling efficiency.* Managed and maintained GCP and OnPrem Kubernetes clusters with infrastructure as code (IaC) using Terraform and ArgoCD for deployment and lifecycle management. Oversaw multiple Java applications, Go-based servers, and data platform consumers, ensuring high availability, efficient scaling, and robust platform performance.* Self-managed Prometheus and transitioned to a federated architecture for enhanced scalability and resilience. Implemented Loki for log aggregation and integrated code profiling using Grafana Pyroscope, and Pixie to optimize performance and monitoring across services.
