Graduate Research Assistant

• Sep 2021 - Present

- Developed cutting-edge machine learning techniques, including graph neural networks, attention-based and transformer-based models, to improve histopathology image cancer classification as a multiple instance learning method with ResNet and Vision Transformer, resulting in a 6.88% increase in AUC. - Designed and built a machine learning pipeline to preprocess giga-pixel images, extract embeddings, and create a graph based on distances for histopathology image analysis. - Applied transfer learning and domain adaptation techniques to H&E histopathology cancer classification, including color normalization and domain adversarial neural networks. - Conducted outcome prediction analysis using tissue images and clinical data to identify key prognostic factors. - Employed multi-modal analysis techniques to estimate other stains with H&E using random-forest and autoencoders to improve prostate grading by deep learning models Show less

Graduate Teaching Assistant

• Sep 2022 - Dec 2022

- CPSC 422 Intelligent Systems: Invigilating and grading exams and assignments.- CPEN 400D Topics in Computer Engineering - DEEP LEARNING: Answering questions in the class, invigilating and grading exams.

Graduate Academic Assistant

• Sep 2021 - Aug 2022

Research Assistant

• Feb 2020 - Apr 2021

- Developed a system for controlling the formation of autonomous vehicles in a maze with dynamic obstacles. - Implemented model predictive control based on convex optimization to control each robot’s kinematics. - Proposed a novel reinforcement learning path-planner to replace traditional potential field and heuristic methods, resulting in more effective and adaptive navigation strategies for the robots using Matlab Simulink. - Developed a system for controlling the formation of autonomous vehicles in a maze with dynamic obstacles. - Implemented model predictive control based on convex optimization to control each robot’s kinematics. - Proposed a novel reinforcement learning path-planner to replace traditional potential field and heuristic methods, resulting in more effective and adaptive navigation strategies for the robots using Matlab Simulink.

Software Developer Intern

• Jun 2019 - Sep 2019

- Surveyed prior state-of-the-art and planned future steps on how to apply and improve them in our task. - Developed low-priced, multi-protocols IR-remote control for televisions and air-conditioners using C in ARM (STM32F030K6) microcontroller which is commercialized as ”Luxin Remote Controller”. - Surveyed prior state-of-the-art and planned future steps on how to apply and improve them in our task. - Developed low-priced, multi-protocols IR-remote control for televisions and air-conditioners using C in ARM (STM32F030K6) microcontroller which is commercialized as ”Luxin Remote Controller”.