Chia-Kai (Kai) Yeh

I’m Kai, an applied science leader with 10+ years of experience at the intersection of computer vision, 3D graphics, machine learning, and generative AI. I earned my PhD in Computer Science from Northwestern University’s Computational Photography Lab, where I specialized in software-hardware co-design for 3D reconstruction and human digitization.

My career spans startups, research labs, and industry — from Adobe Research and USC ICT to 8i, M³, and now Flawless AI, where I’m Director of Applied Science. I’ve led teams in building advanced systems for volumetric capture, photorealistic facial modeling, and 3D reconstruction from RGB/depth sensors, as well as pioneering GenAI-driven tools for storytelling. At Flawless, I help reimagine filmmaking with AI technologies for expression transfer, visual dubbing, lipsync, and other scalable solutions that bring digital humans to life in production pipelines.

I’m driven by transforming cutting-edge research into real-world impact — leveraging AI, computer vision, and computer graphics to amplify human creativity. My goal is to shape long-term AI strategy, scale applied science teams, and push the boundaries of generative media and digital humans.

Email / GitHub / LinkedIn / CV

Research

I'm interested in computational photography, computer graphics, computer vision, 3D reconstruction and appearance modeling.

	Diagnosing a Paint Disease with Computer Science: The Case of Georgia O’Keeffe Chia-Kai Yeh, Vikas Gupta, Florian Willomitzer, Dale Kronkright, Aggelos Katsaggelos, Marc Walton, Oliver Cossairt American Association for the Advancement of Science Annual Meeting, 2019 Video / Media Coverage / We create a novel hand-held metrology tool detects, maps and monitors destructive micron-sized protrusions resulting from a chemical reaction between the metal ions and fatty acids commonly used as binder in paints in precious works of art. Our system utilize only of a mobile hand held device such as tablets and mobile phone and exploits screen and front camera for Deflectometry-based surface measurements. We demonstrate high quality measurements without the need for an offline calibration procedure. This new tool will help conservators easily and effortlessly map and monitor the strange paint disease and prolong the lives of art.
	PhotoGeode: 3D Faces from 2D Imagery Chia-Kai Yeh, TJ Rodes, Bhautik Joshi, Duygu Ceylan, Sunil Hadap Adobe TechSummit, 2019 Video / Media Coverage / We create an icosahedron multi-view capture enviroment for photorealistic human facial geometry and appearance modeling. We had took our project to showcase at 2019 Adobe Tech summit at San Francisco which we had captured nearly 500 faces on scene and created a high quality 3D and appearance human facial dataset.
	CS-ToF: High-resolution compressive time-of-flight imaging Fengqiang Li, Huaijin Chen, Adithya Pediredla, Chia-Kai Yeh, Kuan He, Ashok Veeraraghavan, Oliver Cossairt OSA Optics Express, 2018 Paper / Use compressive sensing to improve the lateral resolution of the off-the-shelf ToF cameras.
	Shape-from-Shifting: Uncalibrated Photometric Stereo with a Mobile Device Chia-Kai Yeh, Fengqiang Li, Gianluca Pastorelli, Marc Walton, Aggelos K. Katsaggelos, Oliver Cossairt IEEE International Conference on eScience, 2017 Project Website / Paper / A mobile device approach for surface normal and shape recovery in the wild.
	A Streamlined Photometric Stereo Framework for Cultural Heritage Chia-Kai Yeh, Nathan Matsuda, Xiang Huang, Fengqiang Li, Marc Walton, Oliver Cossairt ECCV Workshop, 2016 Project Website / Paper / A streamlined framework of robust 3D acquisition for cultural heritage using both photometric stereo and photogrammetric information.