Research Vision
I work on developing next-generation multi-modal spatio-temporal reasoning systems for understanding and acting in complex environments. My goal is to enable transformative advances in the automation of man-made environments through intelligent multi-agent collaboration, in which mobile robots will work alongside humans with augmented assistance. More specifically, I am interested in developing systems for joint reasoning over 3D maps, digital twins, egocentric and exocentric video feeds, and operational documents, enabling continuous, online execution of environment-level goals.
3D Understanding
Building systems for robust 3D scene understanding and natural language interaction in complex environments. The goal is a foundation model that tokenizes complex 3D scenes and enable spatially grounded reasoning for embodied planning, dialogue, virtual tours, storytelling, etc.
Egocentric Understanding
Developing multimodal VLMs for streaming, egocentric video, where an agent must reason over hours of observation under tight context and latency limits. I focus on online frame sampling, structured memory over long horizons, and proactive reasoning that decides when to ask, tell, or wait.
Multimodal Alignment for Embodied Planning
Aligning 3D maps and digital twins with egocentric and exocentric video into a unified representation. With the aim of deploying onto mobile robots to explore practical usability across man-made infrastructure such as construction, logistics, and manufacturing.
About Me
I am a Research Scientist at INSAIT, leading a small research team to develop spatio-temporal reasoning systems that operate over multi-modal data, such as 3D maps, digital twins, and egocentric video streams, aiming to enable collaboration between robots and humans in complex environments. I completed my PhD at the Computer Vision Lab at ETH Zürich, advised by Prof. Luc Van Gool and Dr. Danda Pani Paudel, focusing on 3D reconstruction and visual understanding. Toward the end of my PhD I interned at Meta Reality Labs in Zürich, working on 4D reconstruction. Before that I was a full-time Teaching Assistant at the University of Belgrade, School of Electrical Engineering, where I also earned my MSc and BSc in signal processing and control theory. I am a long-time member of the organizing committee of PSIML, an annual AI summer school in Serbia.
Beyond Research
Mentoring
- – Supervised 10 students (mostly PhD) at INSAIT
- – Supervised 11 MSc/BSc students at ETH Zürich
- – Committee for 12 BSc theses at University of Belgrade
Teaching & Services
- – Research consultant for 2 AI startups
- – TA for Computer Vision course at ETH Zürich
- – TA for 7 courses at University of Belgrade
Community
- – Reviewer for CVPR, ECCV, ICCV, WACV, NeurIPS
- – Organizing annual AI summer school in Serbia
- – Pushing student research initiatives in Serbia