Thang Doan

WORK HISTORY

ABOUT THANG DOAN

I am a Staff Applied Research Scientist at Snorkel AI, working with Vision Large Language Models (VLLMs) for Composed Image Retrieval in the context of online retail. My second area of focus revolves around leveraging LLMs for synthetic dataset generation. Prior to that, I was a Research Scientist at Bosch, where I focused on addressing Open-World problems such as Object Detection and Segmentation within the context of self-driving scenes. Additionally, I specialized in multimodal (Vision-Language) representation learning for lifelong learning. Before joining Bosch, I was a postdoctoral researcher at Mila under the supervision of Joelle Pineau. My research during this period centered on two key streams: developing a theoretical understanding of Catastrophic Forgetting and exploring the application of self-supervised learning in Reinforcement Learning. I earned my Ph.D. from McGill University where I delved into cutting-edge topics such as Generative Adversarial Networks (GANs), with a specific focus on addressing mode collapse problems. In tandem, I explored the realm of deep reinforcement learning, where I tackled exploration challenges by implementing innovative multimodal policies.