• Investigate the memorization of Diffusion models: what samples are memorized the most” and “why do different models trained with different objectives memorize the same samples".• Research into the layer-choice when evaluating memorization. Using PyTorch to train and test different Convolutional Autoencoders and evaluate the CT-score on various Transformer (i.e: ViTs), CNNs. Doing case study for using ViT as image embedding encoder to evaluate memorization using cosine-similarity• Built repo for automate DDPM and GAN sampling and various evaluation with free encoder choice base on provided checkpoints.

• Collaborated with a team of 3 in developing a trip budget planning web app using Java, JavaScript,HTML/CSS.• Integrated Google Cloud’s App Engine and Datastore for efficient data storage and retrieval. Designed encrypted user account systems for sign-up and log-in security• Concentrated on backend development while coordinated with front-end team, crafting functionalities for efficient user, trip, event, and budget data storage, doubled system’s carrying capacity, enhancing overall performance and user experience.• Adopted industry best practices through open-source contributions via GitHub and rigorous code reviews.