Helin Xu | 徐赫临

I am a machine learning engineer at TikTok Inc. in San Jose, California. I received my M.S. in Computer Science from UC San Diego in 2024, and my B.Eng. in Automation from Tsinghua University in 2023.

My research is in embodied AI and robot learning, with an emphasis on generalizable robotic manipulation: 3D perception of articulated objects, reinforcement learning from point-cloud observations, cross-category domain generalization, and large-scale physics simulation for scalable robot training data. I have been very fortunate to be advised by Prof. Xiaolong Wang, Prof. He Wang, Prof. Li Yi, Prof. Hao Su, and Prof. Yebin Liu.

I am a lifelong learner and I aim to build things that work in the real world. I am always open to conversations. If you share similar interests or simply want to chat, feel free to reach out!

Email  /  Twitter /  WeChat  /  Google Scholar  /  GitHub /  LinkedIn

We propose DexArt, a task suite of Dexterous manipulation with Articulated object using point cloud observation. We experiment with extensive benchmark methods that learn category-level manipulation policy on seen objects. We evaluate the policies’ generalizability on a collection of unseen objects, as well as their robustness to camera viewpoint change.

We propose to learn generalizable object perception and manipulation skills via Generalizable and Actionable Parts, and present GAPartNet, a large-scale interactive dataset with rich part annotations.

Dora AI generates websites from a simple natural-language prompt, helping users create websites quickly without coding or design skills. I led the machine learning team, and curated front-end user interface datasets using a deep learning detection model. The product launched on Product Hunt and was ranked #1 Product of the Day, Product of the Week, and Product of the Month.