Mohsen Ghafoorian

Timeline

1. 2005–2012

   AI master's at Sharif

   He studies software engineering at the University of Tehran from 2005 to 2010. He then earns a master's degree in artificial intelligence at Sharif University of Technology in 2012, with a thesis on hierarchical reinforcement learning.

2. 2013–2017

   PhD at Radboud University

   He does a PhD in machine learning at Radboud University from October 2013 to June 2017, working on algorithms that detect and characterize neuro-degenerative diseases. From 2016 to 2017, he is also a visiting scholar at Harvard University.

3. 2016

   Location-sensitive CNNs for brain MRI

   In October 2016, he is first author of a paper that proposes convolutional neural networks with anatomical location information for segmenting white matter hyperintensities in brain MRI scans. On a test set of 46 scans, the best network reaches a Dice score of 0.791, against 0.797 for an independent human observer.

4. 2017

   Deep learning survey for medical imaging

   In February 2017, he is one of nine authors of "A Survey on Deep Learning in Medical Image Analysis", which reviews the major deep learning concepts for medical imaging and summarizes over 300 contributions to the field.

5. 2017–2021

   Deep learning lead at TomTom

   From 2017 to 2021, he works at TomTom in Amsterdam as a senior deep learning R&D engineer, technical team lead and line manager, developing deep-learning methods that generate high-definition maps for autonomous driving. In September 2020, he becomes the TomTom-side director and coordinator of the TomTom–UvA Atlas Lab, a lab with five PhD students that Theo Gevers and Cees Snoek direct on the UvA side.

6. 2018

   EL-GAN for lane detection

   In June 2018, he is first author of EL-GAN, a generative adversarial network framework that discriminates on learned embeddings of both the labels and the prediction at the same time. Tested on the TuSimple lane marking challenge, it simplifies post-processing and crosses the competitive 96% accuracy threshold.

7. 2021–2025

   Extended reality at Qualcomm

   He joins Qualcomm in July 2021 as a staff engineer working on extended reality, and he is promoted to Senior Staff Engineer in December 2023. DG-Recon, a 2023 ICCV paper he coauthors, guides neural 3D scene reconstruction with monocular depth priors and improves on the NeuralRecon baseline while remaining in real time. Until April 2025, he leads Qualcomm's Extended Reality team in Amsterdam, building on-device 3D scene reconstruction and understanding features.

8. Since May 2025

   Efficient video diffusion at Qualcomm

   In May 2025, he starts leading a team on efficient video diffusion models at Qualcomm AI Research, as Senior Staff ML Research Scientist, Technical Team Lead and Manager. In July 2026, he is first author of MobileWan, a video diffusion model built from Wan2.2-5B that runs on a commercial mobile device and generates 5-second 480x832 videos at 16 FPS in 20 seconds; the paper is accepted at NeurIPS 2026 in September 2026.