Singapore
Brian is building world model at AMILabs.
I joined on day one as part of the founding team, nominally as a Research Scientist. In practice, though, the role has been much broader than research alone. A large part of the journey has been about building and scaling things from the ground up - infrastructure, tooling, data pipelines, evaluation systems, and models, often starting from bare metal. This also means the opportunity to connect them together into a system that can actually deliver. Much of the work happens at an unusual scale and often requires diving into the lowest level details, where small decisions can have an outsized impact. Overall, we are turning ideas into reliable models and systems, making things work in real world and create meaningful value for people.
LMMs Lab is affiliated with NTU MMLab and was founded by Prof. Ziwei Liu together with his students, with a shared focus on advancing multimodal models research. It is non-profit and research oriented, bringing together passionate researchers from diverse backgrounds - including both application and theoretical researchers, competitive programming medalists, and frontier hackers - united by a shared commitment to advancing multimodal intelligence. For more details, check our website https://www.lmms-lab.com/
Tiktok@AIIC Co-lead the development of the first generation of Tiktok Visual Foundation Model, this model matched Qwen-2.5-VL's performance and even exceled at video understanding at same parameters size.
with Dr. Chunyuan Li Delivered LLaVA-NeXT, LLaVA-OneVision, LLaVA-Video series.
Research@Microsoft Research, Redmond on multimodal models with Dr. Chunyuan Li. Witnessed LLaVA's early stage development. Delivered projects like Otter/MIMICIT and starts the journey exploring multimodal intelligence.