Shima Shahfar
I am a PhD student at Université de Montréal and Mila, advised by Liam Paull in the Robotics and Embodied AI Lab (REAL). My research is on 3D scene understanding and active perception for robots under uncertainty. Before my PhD, I spent four years building 3D perception systems at Planitar (iGUIDE), where I still work part-time as a research scientist.
Publications
A 3D scene graph whose edges are Bayesian filters, so a robot can predict where objects will be over time.
A gradient-reversal auxiliary loss that disentangles structure from texture in the Swapping Autoencoder.
Experience
2026 – now Machine Learning Research Scientist (part-time), Planitar (iGUIDE). Research on 3D perception for indoor spaces.
2022 – 2026 Machine Learning Engineer, Planitar (iGUIDE). Designed 3D perception systems from 360° images and 2D LiDAR, deployed in iGUIDE’s commercial products.
2022 Research Intern, CAE / Presagis. 3D digital twins of cities from 2D data.
2019 – 2022 M.Sc. researcher, Immersive & Creative Technologies Lab, Concordia University, with Charalambos Poullis. Unsupervised structure-consistent image-to-image translation.
2017 Software Engineering Intern, Rahnema Co. Developed and launched a text-based social networking app for Android and the web.
Service & teaching
2026 – now Organizing team, Women@Mila.
2026 Organizing team, Women Techmakers Montréal, International Women’s Day.
2021 – 2024 Reviewer for CVPR, ICCV and ECCV.
2017 – 2022 Teaching assistant: computer vision, machine learning and AI at Concordia University; data structures, algorithm design and advanced programming at Amirkabir University of Technology, and head TA for introduction to programming.