Kevin J Liang

Dr. Kevin J Liang is a Research Scientist at Meta Platforms, Inc., specializing in Deep Learning, Computer Vision, and 3D Reconstruction. He earned his PhD in Electrical & Computer Engineering from Duke University in 2020, with a dissertation on Deep Automatic Threat Recognition for Airport X-Ray Baggage Screening.

His research focuses include:

• 3D Computer Vision (ICON, Fast3R)
• Few-Shot Learning (Sylph, HyperMix)
• Federated Learning (WAFFLe)
• Object Detection (EgoTracks, Self-Supervised Methods)

Recent publications demonstrate his leadership in Egocentric Vision (Ego-Exo4D) and Transformer Applications (GliTr). He has received numerous awards including the E Bayard Halsted Fellowship (2017) and Summa cum laude (2015), and serves on program committees for major conferences like NeurIPS and CVPR.

As an educator, he developed and taught tutorials for Duke University's Machine Learning School and Coursera courses, covering TensorFlow, PyTorch, and foundational ML concepts for over 600 students.