At Predigle India, I lead eight AI/ML engineers while owning most of the organization’s AI/ML codebase. I build LLM, STT, and OCR frameworks and chatbot orchestration systems that connect AI workflows to backend services.
For EsperPraxi and My Whole Child Pediatrics, I architected a patient and guardian platform spanning appointment booking, registration, AI chat, and document management. I built OCR and LLM extraction with confidence scoring and source-coordinate verification, plus clinical entity resolution with Snowflake masking policies and RBAC.
I redesigned an Airflow ETL pipeline handling a 2.1M-row full load, reducing runtime from 4–5 hours to under 11 minutes. The work combined threaded table processing, incremental watermarks, streamed PostgreSQL cursors, and MongoDB indexes.
Previously, at Rajasthan Royals and Blenheim Chalcot, I built computer-vision pipelines for cricket analysis. I used detection, segmentation, and homography-based reconstruction to track ball trajectories and produce tactical analytics for coaching staff.
