At ICRA Analytics, I build regulated data platforms and deployed an air-gapped LLM assistant that converts plain-English questions into SQL without data leaving the environment. I also designed ingestion and orchestration services handling roughly 5 TB within a six-hour regulatory SLA window with zero downtime, reducing infrastructure cost by 40%.
I've improved Spark pipelines from eight hours to about 90 minutes, built entity resolution across 60 million customer records, and created self-healing Spark clusters for air-gapped infrastructure. Earlier, I delivered bank regulatory reporting, built distributed Scala and PySpark ETL pipelines, and developed a real-time market-data platform with WebSocket streaming and OHLC aggregation.
