I am Michael Firman, a Research Scientist at Third Dimension AI, where I lead a team to develop and ship 3D reconstruction and world models for high-fidelity simulation of the real world. These enable rigorous validation of autonomous systems in complex, real-world environments.
Previously, I was a Senior Staff Research Scientist at Niantic Spatial, where I led research and engineering to bring advanced 3D perception to production, including depth estimation, 3D reconstruction, semantics, AR occlusion, and the infrastructure and evaluation systems needed for large-scale deployment.
Before Niantic, I was a postdoctoral researcher on UCL’s Engage project, making machine learning tools accessible to scientists. I collaborated with Prof. Mike Terry (University of Waterloo), Dr. Gabriel Brostow (UCL), and Prof. Kate Jones (UCL). I also worked at the National Institute of Informatics, Tokyo with Prof. Akihiro Sugimoto.
During My PhD, supervised by Dr. Simon Julier and Dr. Jan Boehm, I developed methods to complete full 3D scenes from single depth images.
I have published at leading venues including CVPR, ICCV, IROS, and ECCV, and have served as a reviewer for CVPR, ECCV, ICCV, IROS, BMVC, ICRA, IJCV, CVIU, 3DV, and ISMAR. I received CVPR’s Outstanding Reviewer award in 2018, 2020, 2022, and 2025.
