sandeep vvs

At Impetus Technologies, I built an Agent-as-a-Service platform by developing agent memory, deployment infra, a2a , mcp , ci/cd, registry, observability, tracing adopted by 20 teams , cutting deployment from days to under 10 minutes, plus a LangGraph LLM RAG chatbot and. I developed ML/LLM monitoring for 40+ production models and 20K+ traces with automated alerts. Earlier, I built PySpark/Airflow pipelines, migrated Redshift data to Palantir Foundry, and automated CI/CD on AWS.

Developed Pyspark applications in a Big Data Hadoop architecture for Deutsche Bank and acted as an point of contact for the offshore team. Optimized long-running Spark jobs through techniques like column pruning and caching, reducing runtime. Designed process flows for additional features and new ETL pipelines, enabling parallel processing in Spark. Wrote unit test scripts with high coverage and improved efficiency by over 50%. Utilized Control-M for job scheduling and Bitbucket for CI/CD.