At Eternal Robotics, I architect distributed inference pipelines for real-time industrial computer vision, supporting 60+ concurrent video streams at 15 FPS per stream.
I engineered ANPR systems using OCR and object detection for vehicle lifecycle tracking across Mahindra and Maruti service centers. I also built production vision pipelines that achieved 98% verification accuracy and reduced manual inspection effort by 80–90%.
I’m leading the design, development, deployment, and optimization of 10 production-grade computer vision applications, while mentoring engineers on system architecture and production deployment. I’ve also improved model robustness through dataset refinement and Grad-CAM-based failure analysis, reducing false detections by 70%.
In my 3D vehicle damage assessment project, I developed a cloud-based pipeline that reconstructs vehicles from LiDAR-enabled iPad captures and compares point clouds with CAD models. I also built a document understanding application for PDF retrieval and question answering.
