Tony Ho

About Me I recently graduated from Simon Fraser University's Master of Professional Computer Science program, specializing in machine learning and computer vision. With a strong foundation in AI and a background as a former game developer, I bring a blend of creativity and technical expertise to my work. You'll find a selection of my projects that showcase my passion for innovation and my proficiency in various areas of machine learning. I am eager to apply my skills to solve complex problems, contribute to cutting-edge developments, and create innovative applications.

Connect4 Action AI This project trains a Faster-RCNN model to detect the state of a Connect4 board and exports it as a PyTorch Mobile model. This model performs inference on an Android smartphone and displays identified objects right on the screen. The inferred state of the board is passed to a min-max algorithm to recommend the best move for the current player.

Amazon Warehouse Classifier This project uses AWS Sagemaker to train image classification models. A ResNet34 model is trained as a benchmark, using data distributed parallelism. A ViT model is trained to investigate inherit advantages of transformer models and compare.

Emoji Prediction This project fine-tunes a DistilBERT model on tweets to predict an appropriate emoji. Gathers real tweets using the Twitter API and filters out irrelevant tweets to build the dataset. Uses the Huggingface transformers package to implement the DistilBERT model.

• DistilBERT
• 🤗Transformers
• Twitter API
• Text Classification
• NLP

Car Avoidance This project trains a car agent to avoid obstacles in increasingly complex environments. Using Unity3D and the MLAgents package, the agent is trained using the Proximal Policy Optimization algorithm in parallel 3D environments.

• Unity3D
• MLAgents
• Proximal Policy Optimization
• Obstacle Avoidance
• Deep Reinforcement Learning