At Rakuten India, I rebuilt an event-processing pipeline in Go and Cassandra from 36.5 to 3,884 QPS/thread, reducing a 60M-event batch from 45 hours to 26 minutes. I also put a Kafka-based notification service into production to transform and route 14B+ identity events monthly.
I build production AI systems including Zuno, an enterprise document intelligence platform using Azure OpenAI GPT-4o, hybrid search, n8n, React, and Next.js. My work spans RAG agents, observability, secure multi-tenant platforms, and explainable medical AI research published through IEEE and ETASR.
