JOSHUA TING
climate – data – photography
PORTFOLIO.
WORK 01
Satellite Imagery Segmentation with Convolutional Neural Networks
Predicting land cover types using remote sensing data with computer vision. Used Convolutional Neural Networks to segment satellite imagery into different land type classes.
WORK 02
AI Plays Ms. Pac-Man
Trained a reinforcement learning agent to play the Atari 2600 game of Ms. Pac-Man. Used a Dueling DQN architecture to confirm scientific research performance and results. Built a web-app to live stream gameplay in real-time.
WORK 03
Woby’s Spooky Tales
Human-machine interactive storytelling engine fine tuned for scary/horror stories. Trained with GPT architecture variants on custom corpus sourced from Reddit for the downstream task of text generation. Integrated final model into a chatbot (Woby) front end for human-machine interface.
WORK 04
Lunar Landscape Image Segmentation
Using custom U-Net to compare against several pre-trained backbone U-Net architectures to segment images of lunar landscapes into various classes.
Joshua Ting 