First-year Engineering Science, University of Toronto
Hello, my name is Jason! I'm a first year Engineering Science student at the University of Toronto interested in robotics, autonomous vehicles, and machine learning!
Being inspired by the potential for self-driving cars to eliminate human-error-induced collisions, I would love to work in the self-driving cars industry! So, I worked through the Self-Driving Cars specialization by Steve Waslander and am planning to join the aUToronto self-driving cars design team at the University of Toronto.
I completed Andrew Ng's Deep Learning Specialization and am currently exploring machine learning by building projects such as a U-Net image segmentation model.
EXPERIENCE 01
Used Python and Flask to create a back-end for fact-checking a user's writing using transformers. Used TypeScript and React.js to redesign the front-end of a software application in collaboration with Figma graphic designers. Coordinated with a team of software developers through GitHub in a dynamic start-up setting.
Projects
A running list of my personal projects with their GitHub pages attached!
PROJECT 01
I used Python and TensorFlow to design and train an 8 million parameter U-Net image semantic segmentation model for self-driving cars. The model was trained on a CARLA-generated dataset using Keras with NVIDIA Tesla P100 GPU acceleration. The neural network produces 23-class pixel-wise segmentation masks and the predictions are visualized using color overlays on Toronto driving footage. Furthermore, I used Hough Transforms to perform lane detection using classical computer vision in OpenCV.
PROJECT 02
Implemented an error-state extended Kalman filter for state estimation for an autonomous vehicle. Used IMU dead-reckoning for prediction and corrected the state estimations with sample LiDAR and GNSS measurements.
PROJECT 03
Used Python to create a vehicle controller to guide a simulated vehicle through waypoints in CARLA. Implemented a PID controller for longitudinal control and the Stanley controller for lateral control. Built a finite state machine to navigate stop sign intersections and follow leading cars in CARLA. Created a conformal lattice planner to avoid obstacles in CARLA while traversing the reference trajectory using a cubic spiral path generator and circle-based collision checker.
PROJECT 04
Created a software in Java that can analyze and evaluate chess positions and play against the user by searching chess positions into the future. The algorithm can search and evaluate over 260,000 positions per second.