At Krishaka, I own architecture and delivery for production AI systems across LLM applications, edge inference, and robotics workflows. I build backend services, deployment pipelines, and monitoring from prototype through field use.
I developed production LLM/RAG backbones, fallback routing, and observability for internal and customer-facing workflows. I also built UWB optimization and failure-prediction systems that enabled sub-50 ms inference across 200+ nodes while reducing coverage gaps by 40%.
Previously at Navi Technologies, I improved the Querybuilder text-to-SQL assistant by diagnosing retrieval, schema, and table-ranking failures, reducing unsolved queries by about 20%. I also built PySpark and Databricks risk-monitoring pipelines and an XGBoost risk-tier model that improved KS by 33%.
I bring hands-on experience with LLM systems, reinforcement learning, computer vision, robotics, and MLOps. My work includes an awarded patent for a Computer Vision and UWB-based Autonomous Navigation System.
