ML/AI Uber TL (ComputerVision focus) of the Cross-Meta Integrity AI org (80+ ML/AI engineers) leading team of strong Research Scientists building state-of-the-art visual understanding models, such as representation learning, object detection, and Multimodal LLMs.Well-known for building the computer vision circle in the integrity community; strong track records of delivering great business outcomes by leading research-to-production and 0-1 deep learning computer vision innovations across organizations with deep impact on the work of dozens of teams and 100+ ML engineers.

I work on machine learning algorithms and systems to generate similarity-preserving hash codes for large-scale video retrieval and clustering. My algorithm and system pipeline process tens of millions of videos uploaded to Facebook everyday.

I worked on the massively-distributed and highly-parallel columnar database system. Specifically, I applied robust optimization theory to database physical design. I implemented a robust physical design algorithm and integrate it with Vertica's existing DBDesigner. The algorithm is able to maintain stable performance against perturbations in query workloads.

Research on automatic object recognition using template matching algorithms. Build 3D model of scenes to obtain object templates of different scales and views. Design algorithms to automatically pick the best template to match with target object.Participate in the implementation of a system prototype using C++, MFC and OpenGL.

Research on biologically-inspired machine learning algorithms for object classification.Lead a group of two to build a system prototype using MFC with the researched object classification algorithm as the core.Design a topology image database and write C++ and MATLAB algorithms to generate topological transformation of images of basic shapes.