CV

Co-founder & CTO

Outrun Robotics, Inc

Building the deployment layer for accelerating wide-spread adoption of robots.

• Architected and developed the perception, compute, and vision-monitoring systems for the remote autonomy stack.
• Deployed two robotic systems to a factory production environment in under six months with $500k in funding, then scaled to two additional production facilities, reaching pilot performance within one week.
• Developed a HIL-RL based post-training pipeline to make large behavior models performant and reliable.

Staff Roboticist

Tacta Systems, Inc

• Engineered sensorimotor policies for dexterous multi-fingered manipulation using tactile sensors.
• Designed and maintained end-to-end robotics software, building key demonstrations for a Series-A raise.

Senior Robotics Software Engineer

Dexterity, Inc

• Designed, developed, and deployed a grasp prediction model optimizing object selection and grasp pose from overhead RGBD camera data.
• Developed an IoU-based tracker that increased picking speed and improved handling of dynamic scenes, helping the parcel sortation product reach 2,000 PPH across 75+ robots.
• Built and maintained a regression-testing framework for perception pipelines to catch performance regressions across product lines.

Senior Software Engineer

Aurora Innovation

Continued Uber ATG projects post-acquisition, focusing on trajectory forecasting using graph neural networks.

Senior Autonomy Engineer: Deep Learning for Perception & Prediction

Uber ATG (Advanced Technologies Group)

• Developed an efficient 2D multi-class detection network (RFCN) with state-of-the-art metrics, improved further via hard example mining.
• Built a novel sensor-fusion (image + lidar) 3D detection model that became a production model running on 50+ self-driving vehicles; resulted in a patent and a CVPR Workshop 2019 publication.
• Extended the sensor-fusion approach to joint perception and prediction (MultiXNet), submitted to ICRA 2021.
• Mentored two interns; one internship led to a NeurIPS 2019 publication.

Machine Learning Engineer

NVIDIA

• Contributed to a
• Built the image-segmentation demo showcased by NVIDIA at CES 2016.
• Worked on end-to-end driving in simulation (TORCS) using deep reinforcement learning (A3C).

Part-time MS (did not complete)

Stanford University

Computer Science

• Relevant coursework: Introduction to Robotics (CS223A), Advanced Robotic Manipulation (CS327A), Deep Reinforcement Learning (CS224R), Optimal and Learning-based Control (AA203), Sensorimotor Learning for Embodied Agents (CS381).
• Independent research on dexterous manipulation, advised by Prof. Oussama Khatib.

MS

New York University

Data Science

• Relevant coursework: Deep Learning, Statistical NLP, Inference & Representation (Probabilistic Graphical Models), Numerical Optimization, GPU Computing, MCMC Methods.

Bachelor of Engineering

R V College of Engineering

Electronics and Communications

Multi-View Fusion of Sensor Data for Improved Perception and Prediction in Autonomous Driving

WACV 2022

Improving Movement Prediction of Traffic Actors using Off-road Loss and Bias Mitigation

NeurIPS Workshop on Machine Learning for Autonomous Driving (ML4AD), 2019

Sensor Fusion for Joint 3D Object Detection and Semantic Segmentation

IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2019

Convolution Neural Networks for representing sentences/documents

Master's Project, New York University, 2015

Temporal Analysis of Language through Neural Language Models

ACL 2014

Limca Book of Records – National Record

Awarded for building & launching India’s smallest satellite as part of Team STUDSAT.

Consumer Technology Scholarship

Consumer Reports

Awarded from Fall 2013 to Spring 2015.