I'm

Implemented 7+ Safety Samples for NVIDIA's TensorRT's Drive OS 6.0.6

Contributed 4000+ lines of Safety certified code to TensorRT in C++.

Implemented motion planning algorithms for a fleet of autonomous UAVs with sensors onboard to optimize multiple target recognition and localization.

Implemented auction based bidding and planning for motion planning of multiple UAVs.

Contributed to the development, testing and documentation of MagikEye-SDK to extend 3D depth-sensor’s usage on multiple platforms viz. ROS, PCL, Open3D, and Unity; Developed production level, low-latency Python and C++ ROS and ROS2 packages, performed unit-testing, and documented their usage as a part of SDK.

Quantified blurriness and overexposure utilizing classical CV algorithms in scikit-learn for infra-red image classification and fps improvement; deployed optimized machine learning pipeline in C++ to the propriety depth-detection pipeline.

Implemented custom RANSAC for tray-detection and DBSCAN clustering algorithm for object-measurements from 3D Point-cloud.

Fire-fighting drones – Payload drop Module simulation Gazebo | ROS | Ardupilot

• Implemented drop location calculation algorithm by considering payload parameters and wind resistance for fire-fighting VTOL using ardupilot stack; achieved repeatability and accuracy of +/-0.5m at UAV velocity of 10m/s and altitude of 50m.

• Achieved repeatability and accuracy of 0.5m from the desired target location at UAV velocity of 10m/s and flying altitude of 50m.

Bydrone project - Precision Landing for UAVs C++|OpenCV |Gazebo | ROS | Pixhawk| PX4 flight stack

• Estimated extrinsic parameters of the camera mounted on UAV using the Perspective-n-Point algorithm with waypoints detected from April tags.

• Implemented UAV pose estimation and position control algorithm to land the UAV at a static target location.

Multi-Camera Human Tracking for security |Python | OpenCV |Tensorflow |Keras | PIL| Numpy

• Implemented a combined HSV histogram-based object tracking and a fully convolutional Siamese network for online training and tracking of pedestrians passing through multiple cameras.

• Improved tracking accuracy by handling occlusion and retracking in <>70% cases.

Robotic transmission assembly |Universal robot programming |Optoforce F/T sensor

• Developed an algorithm for dowel insertion on crankcase base and crankcase insertion through shafts using force and torque feedback along 3-axis with Optoforce F/T sensor.

• Calculated offsets using axis-transformation between 2 differently-oriented universal robots and implemented synchronized bolt insertion and bolt tightening by 2 robots.

• Programmed for sequencing logic to optimize cycle time ensuring collision free operation of robots.

IndustrialCam – Assembly inspection and Robot Guidance | Python | Kivy | OpenCV |

• Implemented the OpenCV features viz. line, circle and corner detection for assembly inspections and implemented object recognition using ORB feature matcher to estimate object pose for robot guidance.

• Programmed robust algorithms on a low-cost web-camera (5x cheaper) as an alternative to commercial smart cameras and saved ~$4500 per deployment.

Camshaft load- unload automation |Python |OpenCV |Universal Robots| Raspberry Pi

• Estimated the angles of lobes of camshaft with HSV colour space based contour extraction using OpenCV.

• Logic development for gripping and camshaft aligning according to the angle of lobes for accurate placement in fixture of Bearing Press Machine.

Vehicle inventory management system |UWB Beacon | Python| MS Excel - Pivot tables & Macros

• Conceptualized and implemented a novel approach to perform indoor navigation using UWB Beacons to track labours and thereby estimating Vehicle inventory using feedback from Android App.

• Performed data-analysis to analyze live dispatch trends dealer-wise, SKU-wise and region wise and developed logic to dynamically reallocate the incoming SKU’s based on these trends.

• Developed a mathematical model to modify storage of vehicles for optimization of travel times for vehicle transfer thereby reducing 6 manpower/day.

Studied the various stresses developed in piping using CAESAR 2.0.

Calculated the thickness for safer design of pipes and material selection for piping.

Studied the parameters for organizing the layout of the piping plant and modelled the layout using MicroStation.