Tejaswi Kasarla

Tejaswi Kasarla is a Researcher at the VIS Lab, University of Amsterdam, and a Research Scientist Intern at Meta FAIR in Paris. Their work intersects non-Euclidean representation learning (hyperspherical and hyperbolic deep learning) for open-world understanding and multimodal foundation models.

• Education: Fourth-year PhD candidate at University of Amsterdam (since Oct 2021).
• Research Experience: Intern at Bosch (Jun 2018–Oct 2018), then Computer Vision Researcher at Bosch (May 2019–Jun 2022). Currently organizing community initiatives like the Women in Computer Vision (WiCV) Workshop at CVPR 2021–2022.
• Teaching: Teaching Assistant for Applied Machine Learning (Nov 2022, Nov 2021).

Research Focus: Non-Euclidean deep learning (hyperspherical/hyperbolic), multimodal foundation models, active learning, and uncertainty quantification in computer vision. Recent work explores hyperbolic safety-aware vision-language models (CVPR 2025 Highlight) and lightweight uncertainty quantification for terrain traversability (ICRA 2024).

Publications: Their research appears in top venues like CVPR, NeurIPS, WACV, and workshops (ECCV Beyond Euclidean, ICRA Resilient Off-road Autonomy). Key themes include class separation, active learning, and hyperbolic embeddings.

Community Contributions: Organized WiCV Workshops at CVPR 2021–2022, served as a board member since 2022. Peer-reviewed for ICCV, NeurIPS, ICLR, and WiCV/NeurIPS workshops.

Interests Beyond Research: Photography and specialty coffee (home barista).