I'm an AI researcher with a long-running obsession: how do machines learn to see, and more specifically, how do they reason about the geometry of the world around them. The shapes, the depth, the structure that tells us where things are and how they relate to each other.
My PhD at the University of Ottawa was an early attempt to answer that. Working with the VIVA Research Lab under Robert Laganière, I was trying to reconstruct living, moving scenes in 3D from a single RGB-D camera, turning depth and colour into volumetric video you could step inside in VR. It meant weaving together computer vision, custom ML models, and geometric reconstruction into something that felt, briefly, like magic.
At MasterpieceX, that question shifted. Rather than reconstructing reality, I turned to reconstructing imagination, building tools that let artists and creatives bring what's in their heads into the digital world. As Head of R&D, the challenge became applying the same geometry and vision techniques to a space that doesn't follow the rules of the physical one.
Outside of that, I dabble in game dev, graphics, and deep dives to questions about space, geology, and how biological systems organise themselves. The Projects section is where the tinkering lives. Based in Ottawa, Canada.
Python · PyTorch · Machine Learning · JavaScript · Three.js · WebGL · CUDA · TypeScript · C++ · C# · Unity3D · Unreal Engine · Agentic Frameworks
Blog
Notes on AI, research, 3D graphics, and things I find interesting.
Loading…
Projects
Papers I've written and presented, experiments in 3D graphics and interactive computing, random things I've done.
- Spearheaded 60+ R&D initiatives to modernize the 3D content creation pipeline by integrating advanced geometry processing and generative AI, reducing bottlenecks, empowering user creativity, and accelerating iteration cycles.
- Conceptualized and launched MasterpieceX, a text/image driven generative AI tool for end-to-end 3D asset creation including animation with over 178k users, addressing a critical need for accessible 3D asset creation.
- Designed and implemented modular agentic-based 3D content workflows integrating human oversight to boost creative output speed and quality while maintaining precise creative vision.
- Engineered advanced geometry processing algorithms (auto-UV unwrapping, remeshing, mesh booleans, voxelization, skeletonization, auto-rig & skin) resulting in over 80% reduction in time-to-creative output.
- Addressed multi-platform compatibility challenges by refining data structures and algorithms to enable seamless deployment across cloud, desktop, and Android (Quest 2), resulting in enhanced system reliability and user reach.
- Developed cutting-edge immersive VR applications (mesh sculpting, texture painting, rigging & skinning, 3D animation creation and editing) for desktop PC and Quest 2.
- Architected, deployed and maintained 25+ containerized solutions resulting in streamlined operations.
- Defined and executed the strategic vision for R&D, aligning technology roadmaps with business goals to drive innovation with cross-functional teams.
- Mentored and fostered the growth of 12 developers resulting in zero turnover.
- Designed and implemented a novel framework for creating immersive VR experiences from a single RGB-D sensor.
- Developed novel methods for background/foreground separation, dynamic 3D scene reconstruction, video summarization, and ML models for omni-directional depth map generation and 3D shape completion.
- Teaching: Discrete Structures, Introduction to Computing I (Java & Python).
- Implemented surface parameterization algorithms for unfolding 3D meshes of tunnels for geological analysis.
- Design and analysis of mesh processing algorithms to extract geological information: fracture orientation, surface roughness, and feature detection.
- Teaching: Discrete Mathematics, Abstract Data Types and Algorithms, Introduction to Computer Science, Internet Application Programming.
- Core development of VideoSphere Command, an IP video management software suite.
- Prototyped integrations of VideoSphere SDK with external partners.
- Implemented color management features, enhancements and fixed defects in CorelDRAW Graphics Suite X5.
- Integrated a C# WPF asset manager into the C++ CorelDRAW environment, allowing users to quickly access all media files within CorelDRAW.
- Core development of J9 Java Class Libraries; added regression testing and automated execution environment management.
- Implementation of a novel image segmentation algorithm using the Well-Separated Pair Decomposition data structure.
- Designed and trained a CNN using TensorFlow to predict depth maps from omni-directional stereo (ODS) images.
- Applied the model to ODS videos to enable 6 degrees-of-freedom movement for VR headset users.
- Source code accompanying the paper presented at IEEE VR 2019.
- Developed a clustering algorithm using the principles of gravity.
- Implemented a 2D visualizer as an interactive demo.
- Shuang Xie, Po Kong Lai, Robert Laganiere and Jochen Lang. "Effective convolutional neural network layers in flow estimation for omni-directional images." 3DV 2019, pp. 671–680. IEEE.
- Po Kong Lai, Weizhe Liang and Robert Laganiere. "Additive depth maps, a compact approach for shape completion of single view depth maps." Graphical Models 104 (2019): 101030.
- Po Kong Lai, Shuang Xie, Jochen Lang and Robert Laganière. "Real-time panoramic depth maps from omni-directional stereo images for 6 DOF videos in virtual reality." IEEE VR 2019, pp. 405–412.
- Po Kong Lai and Robert Laganière. "Creating Immersive Virtual Reality Scenes Using a Single RGB-D Camera." ICIAR 2017, pp. 221–230. Springer.
- Po Kong Lai and Claire Samson. "Applications of mesh parameterization and deformation for unwrapping 3D images of rock tunnels." Tunnelling and Underground Space Technology 58 (2016): 109–119.
- Po Kong Lai, Marc Décombas, Kelvin Moutet and Robert Laganiere. "Video summarization of surveillance cameras." AVSS 2016, pp. 286–294. IEEE.
- Po Kong Lai, Claire Samson and Prosenjit Bose. "Visual enhancement of 3D images of rock faces for fracture mapping." International Journal of Rock Mechanics and Mining Sciences 72 (2014): 325–335.
- Po Kong Lai, Claire Samson and Prosenjit Bose. "Surface roughness of rock faces through the curvature of triangulated meshes." Computers & Geosciences 70 (2014): 229–238.
- Tara McLeod et al. "Using video acquired from an unmanned aerial vehicle (UAV) to measure fracture orientation in an open-pit mine." Geomatica 67, no. 3 (2013): 173–180.
- Alexandre Boivin, Po Kong Lai et al. "Electromagnetic induction sounding and 3D laser imaging in support of a Mars methane analogue mission." Planetary and Space Science 82 (2013): 27–33.