Research Interest

-- I solve machine perception problems using deep learning and statistical estimation tools--

• Deep learning: Experience in developing real-time deep learning models solving a variety of tasks such as sensor data fusion, motion/trajectory and behavior prediction. Models run in real-time, with low memory footprint.

• Deep Graph Models: Explore and manipulate deep graph models and training mechanisms to come up with more efficient models with the goal of real-time, memory optimized deep models.

• Motion /Trajectory, Action and Behavior Prediction/ Forecasting: My research focuses on investigating real-time deep models for multiple agent’s motion, action and trajectory forecasting that accounts for the uncertainty in the predictions. I attempt to model the interaction between the agents using deep graph nets. Research new metrics that better evaluate the performance of such models.

• Localization: Industrial experience in non-visual localization using classical and data driven approaches.

• Sensors: Experience with IMUs, magnetometers and ultrasonic developing heading and tracking systems. Combining both physics and data-driven solutions.

• Scholar: 600+ citations, publications at top venues CVPR, ECCV and AC/Reviewer for top venues. Co-supervise PhD students working on multidisciplinary research.

• Software Engineering: Python, Deep learning libraries, C++, Matlab