- Developing and maintaining a Python-based internal library to manage machine learning models for face detection, landmark detection, and pose estimation. This involves creating a robust ML pipeline with an emphasis on reproducibility.- Conducting research to identify faster and more accurate models for face detection, landmark detection, and pose estimation, and integrating these models into our library.- Leading synthetic dataset generation using diffusion models, ControlNets, 3DMM models, LORA, IP Adapters etc.

Working on the intersection of classical and deep methods for Spatial AI.

- Dense visual SLAM systems - Visual Inertial Odometry- Object detection/segmentation - Unsupervised object categorization - Visual attention

- Lead a team to build and ship inference APIs for recipe prediction(via food detection). - Worked on various ML models for classification, detection and segmentation.- Built data pipeline (labeling to inference APIs), internal framework for managing ML training and serving jobs.

- Implemented lane detection, road segmentation and vehicle detection algorithms for autonomous car project.- Developed driver monitoring system to alert the driver for possible distraction or drowsiness.- Responsible for data pipeline and administration.