• Developed perception pipeline with stereo images as input, generated dense point clouds of the vines using RAFT Stereo; point cloud registration through the iterative closest point algorithm to capture the complete scene and to identify cut points in the generated 3D model• Integrated this pipeline with ROS enabling real-time intra-system communication• Designed a deep stereo network (RAFT) retraining pipeline to improve results on the relevant dataset; used instant-ngp NeRF to generate the ground truth disparity/depth by simulating novel stereo views

TA for 16-299: Introduction to Feedback Control Systems

Spearheaded the development of depth estimation frameworks optimized for large field-of-view (FoV) camerasAchieved state-of-the-art performance on multiple fisheye and 360-degree datasets, demonstrating superior zero-shot generalization

At Bosch Research, I'm working on creating a SOTA model for metric monocular depth estimation (MDE) using deep learning for autonomous vehicles.

Teaching Assistant for the course 24-451 Feedback Control Systems

Designed stereo camera infrastructure to mount it on 6-axis serial manipulators like KUKA Agilus and UR5 to enable them to pick, orient and place any random objectDeveloped a robot calibration system to deploy it in a new facility; devised the algorithm to find the 3D coordinates of the object of interest with respect to the robot base as viewed by cameras and verified the results experimentallyDesigned robot pedestal to maintain the robot position during torque generation while part tending operationsDesigned the PCB enclosure to mount it to the stereo camera system with considerations of thermodynamics in the PCB

Handled the techno-commercial aspects of mechanical, piping, and electrical and instrumentation design in different stages of projects for Beaumont RefineryCompleted a major project of 30 MUSD by developing design improvements in terms of cost, schedule, and feasibilityDeveloped optimized cost estimates for refinery projects by carefully studying the single-line diagrams, piping and instrumentation diagrams, and CAD models of the site; cut down the CAPEX by 16% from the expected amountDid a value capture of 364 kUSD in the flare pilots installation project of 5 MUSD at the Beaumont Polyethylene plant by optimizing the project scheduleDeveloped metrics concerning manhour rate, engineering change orders, and schedule compliance to evaluate engineering contractor’s performance for US sites; data would be used in the contracting strategy of the future projects

Reduced average method time by 34% by designing fixtures for electric locomotive parts for maintenance activitiesDesigned a pit for Bogie overhaul considering the ergonomically safe and more efficient operationOptimized the process of moving bogie axles in the depot by designing a faster ground transportation solution, freeing up cranes for other tasksReduced the manpower requirement by conducting method time analysis and task sequencing for preventive maintenance activitiesDeveloped and tracked safety, quality, cost, and delivery KPIs for insights into industrial engineering department performance to identify the bottlenecks if any, and resolve them

Researched methods to prevent late diagnosis of coronary artery disease, AtherosclerosisDesigned the stenosed artery after literature review and used shear thinning and shear thickening model of blood for transport analysisQuantified wall shear stresses and blood flow parameters like pressure and velocity in the stenosed artery due to plaque formation; the results showed the possibility of relating the blood pressure to the wall shear stresses and blood flow behavior for early detection of Atherosclerosis, complementing angiography