Hi! I am an AI research scientist at Meta Superintelligence Labs working on video and image generation, as reported by The Information. I built a SoTA video editing model, Muse Image and Muse Video. Before that, I worked in OpenAI on image generation. I built GPT Image 1.0 and GPT Image 1.5. Before that, I worked at Google Deepmind on multimodal! Below are more details about public projects I have been working on.

My work at OpenAI is featured by a live stream with Sam Altman, Gabriel Goh, Prafulla Dhariwal, Allan Jabri, and Mengchao Zhong to introduce and demo 4o image generation.

Over 130 million users generated more than 700 million images within the first 10 days of launch, sparked by a viral trend around generating images in the Ghibli style. The trend spread widely, reaching celebrities and even official government accounts on X. Feel free to check out my demo showcasing image upload and style transfer using 4o image generation.

Before this, I worked in Google Deepmind on image perception. I built a model that achieves hair-strand-level segmentation accuracy. It has been integrated in various Google Products (Shopping, Slides, Photos, etc) for billions of users.

Google in Mountain View was my first full-time position after graduation. My education spans three continents, including studies in Australia, Canada, and China, focusing on general representation learning in machine learning and deep learning. Feel free to check out my Google Scholar which mostly reflect what I did during my education period of time.

Physically, I live in bay area. Virtually, you can find me on X.