• Founded a startup using AI and autonomous drones to plant trees, earning seed funding from the University of Toronto Entrepreneurship Hatchery• Created a deep learning image segmentation network for drones to identify land types with 85% success rate• Trained a deep learning object detection model (YOLOv10) to identify trees with 70% precision and 80% recall

• Conducted research on reinforcement learning-based model predictive control (MPC) for UAV trajectory tracking• Developed a quadrotor MPC simulation using Casadi to control a quaternion-based model of the quadrotor• Developed a reinforcement learning algorithm to learn MPC parameters, increasing trajectory tracking performance of MPC algorithms

• Designed and built a cutting-edge autonomous drone for nuclear safety inspections with 90% success rate in mock scenarios. Equipped with CV object detection (YOLOv5), visual SLAM for accurate localization, RRT* for obstacle-free path planning, and precise waypoint motion planning for navigation• Conducted extensive research on diverse nonlinear control methods, systematically comparing them to classical PID control for quadrotor UAVs through simulation using Python• Developed… Show more • Designed and built a cutting-edge autonomous drone for nuclear safety inspections with 90% success rate in mock scenarios. Equipped with CV object detection (YOLOv5), visual SLAM for accurate localization, RRT* for obstacle-free path planning, and precise waypoint motion planning for navigation• Conducted extensive research on diverse nonlinear control methods, systematically comparing them to classical PID control for quadrotor UAVs through simulation using Python• Developed fixed-wing and multirotor UAV simulations in Simulink, meticulously emulating aircraft performance• Programmed and raced Crazyflie drones through gates while avoiding obstacles after simulating drones in a PyBullet simulation, achieving the third fastest time out of ten teams• Trained a drone during flight to avoid obstacles using reinforcement learning in AirSim with a 92% avoidance rate Show less

• Developed an open-source image-guided surgical robotics system, enabling a robot to navigate waypoints specified by users through medical software or real-world ArUco markers, advancing surgical robotics in an underexplored field• Successfully integrated a Franka Emika Panda robot with 3D Slicer medical imaging software using ROS2• Implemented both machine learning-based elastic registration and rigid registration techniques, employing ArUco markers and OpenCV to align coordinate… Show more • Developed an open-source image-guided surgical robotics system, enabling a robot to navigate waypoints specified by users through medical software or real-world ArUco markers, advancing surgical robotics in an underexplored field• Successfully integrated a Franka Emika Panda robot with 3D Slicer medical imaging software using ROS2• Implemented both machine learning-based elastic registration and rigid registration techniques, employing ArUco markers and OpenCV to align coordinate systems between 3D Slicer and the robot, resulting in millimeter accuracy Show less

• Played a key role in aUToronto's win of the SAE/GM AutoDrive Challenge, an annual collegiate autonomous vehicle competition, for three consecutive years, managing localization, mapping, and perception calibration teams of 10 people• Researched state-of-the-art localization methods and developed a semantic localization algorithm in C++ for an autonomous vehicle, utilizing lane lines and traffic lights as markers, achieving first place for the localization task• Conducted calibration of… Show more • Played a key role in aUToronto's win of the SAE/GM AutoDrive Challenge, an annual collegiate autonomous vehicle competition, for three consecutive years, managing localization, mapping, and perception calibration teams of 10 people• Researched state-of-the-art localization methods and developed a semantic localization algorithm in C++ for an autonomous vehicle, utilizing lane lines and traffic lights as markers, achieving first place for the localization task• Conducted calibration of perception system cameras, determining camera intrinsics and extrinsics, and calibrated solid-state lidars using iterative closest point, establishing their transforms, increasing perception accuracy by 30% Show less

• Utilized Simulink for modeling embedded controls software for electric and autonomous vehicle components, including brakes, wheel speed sensor, and IMU, with a focus on diagnostics, contributing to three design teams• Formulated comprehensive test cases in Simulink and validated the functionality of embedded controls software• Generated C code from Simulink models, adapted signal interfaces, ensured successful software builds, did MIL/SIL tests on the model, and uploaded software to… Show more • Utilized Simulink for modeling embedded controls software for electric and autonomous vehicle components, including brakes, wheel speed sensor, and IMU, with a focus on diagnostics, contributing to three design teams• Formulated comprehensive test cases in Simulink and validated the functionality of embedded controls software• Generated C code from Simulink models, adapted signal interfaces, ensured successful software builds, did MIL/SIL tests on the model, and uploaded software to the brake system, handling 10-15 senior engineer level tasks per sprint• Conducted simulations on the test vehicle and controls software using CarSim, GMSimTools, and MDA, leveraging results to pinpoint issues and debug signal outputs for a scrum team of 15 people• Devised an ML model in Python for predicting changes in wheel speed with 87% accuracy within 15% of the speed Show less

• Developed software using ROS with Python and C++ in Linux for the perception, control, planning, mapping, and localization of an autonomous snowblower robot, contributing 20% of the robot’s software stack• Designed a chute planner within the software to dynamically adjust the trajectory of the robot's chute as it moved, ensuring precise snow deposition at specific dump sites on the map while avoiding obstacles in simulation and reality• Implemented and fine-tuned a PID controller to… Show more • Developed software using ROS with Python and C++ in Linux for the perception, control, planning, mapping, and localization of an autonomous snowblower robot, contributing 20% of the robot’s software stack• Designed a chute planner within the software to dynamically adjust the trajectory of the robot's chute as it moved, ensuring precise snow deposition at specific dump sites on the map while avoiding obstacles in simulation and reality• Implemented and fine-tuned a PID controller to enable the robot to accurately follow the desired trajectory• Created a map representation of the robot environment using MapServer• Tested and validated the functionality of the software on the prototype robot for 15 hours per week Show less

• Developed a numerical simulation for tracing light rays and a program to generate realistic computational schlieren images in MATLAB with 85% accuracy to experimental data, contributing to blast wave schlieren research• Tested and validated the functionality of computational schlieren image generation by plotting the results of ray tracing and the intensity distribution of generated images using MATLAB• Wrote a report explaining and documenting simulation code, concepts applied to… Show more • Developed a numerical simulation for tracing light rays and a program to generate realistic computational schlieren images in MATLAB with 85% accuracy to experimental data, contributing to blast wave schlieren research• Tested and validated the functionality of computational schlieren image generation by plotting the results of ray tracing and the intensity distribution of generated images using MATLAB• Wrote a report explaining and documenting simulation code, concepts applied to ray tracing from research papers, computational schlieren image generation, and simulation results Show less

• Designed a pressurant gas diffuser for nitrous gas for a rocket propulsion system using SolidWorks• Designed the cold flow injector plate using SolidWorks and calculations based on launch environment• Participated in cold flow and hot fire testing and designed the test stand for a hybrid rocket• Developed orbital thermal and power simulations in Python for a Cube-Sat satellite

• Set up, operated, and troubleshooted problems with automated assembly machines in an electronic circuit board production line, producing over 20 completed electronic circuit boards per day, improving efficiency by 20%• Manually assembled PCB boards and assemblies, accumulating over 100 hours of hands-on soldering experience with surface-mount, through-hole, and wired assemblies• Visually and electronically inspected 70+ finished PCB assemblies for quality control

• Co-founded two student run companies and gained experience in production, marketing, and sales