Meghasrivardhan Pulakhandam

Architected fault-tolerant REST and async microservices in Python/FastAPI as the ingestion layer for a real-time data platform on AWS across three regions. Built the Kafka Structured Streaming layer with exactly-once semantics, Avro Schema Registry, and consumer group scaling to support sub-100ms p95 latency for fraud detection. Engineered batch processing with Apache Spark/PySpark/Photon, improving job execution time by 35% across 30+ TB/day workloads. Implemented multi-tier Redis caching for LLM inference endpoints (write-through caching and TTL eviction), reducing cold start latency by 60% and lowering database pressure. Containerized and orchestrated services on Docker/Kubernetes (EKS) with HPA; added gRPC for inter-service communication and service discovery. Instrumented end-to-end observability using Prometheus, Grafana, and OpenTelemetry to reduce MTTD by 30%.

Built payment processing services in Java/Spring Boot with REST APIs supporting 10M+ monthly transactions, improving API response times by 35% through performance tuning. Designed async event-driven transaction workflows with Kafka (partitioned topics and consumer group scaling), doubling throughput without SLA degradation. Improved high-frequency transaction read performance at 100M+ record scale using schema redesign, composite indexing, normalization, and Redis caching, reducing critical query latency by 45% and database read load by 50%. Led notification/webhook functionality using GraphQL in Node.js/Express, processing 500K+ events daily and reducing downstream latency by 25% via flexible, schema-driven access. Led AWS cloud migration (EC2, S3, ElastiCache) with Kubernetes orchestration and Jenkins CI/CD, cutting release cycles by 50% and improving deployment success rate to 98%. Implemented observability (Prometheus/Grafana/distributed tracing) and enforced OAuth 2.0/JWT security