KAUSHIK KUMAR Kolar Ravindra Kumar

I’m an AI Engineer and Product Strategist specializing in agentic AI, compound AI systems, and scalable MLOps. I focus on architecting privacy-preserving RAG pipelines and deterministic multi-agent workflows that translate advanced research into measurable business outcomes.

In my recent internship, I architected an end-to-end agentic SDR parsing system that extracted 7+ structured entity types from DDD Service Detail Reports, reducing manual data entry from hours to under 3 minutes per document. I also built strict, auditable agent backend wrappers with check-then-create-or-update logic to constrain agent operations and eliminate unauthorized API access risk, while adding an HITL validation pipeline that maintained accuracy across multi-stage workflows.

Earlier, I delivered backend and performance improvements through microservices—boosting operational efficiency and response times—while strengthening reliability via unit and integration testing practices. I also built real-time computer vision and time-series models, then optimized AI deployments on GCP to reduce latency and scale throughput. My projects reflect my approach to safety and observability: I integrated MCP-backed, secure multi-source retrieval and designed an end-to-end safety system with guardrails and citation validation, supported by evaluation and tracing to continuously improve agent performance.