I’m a Staff Research Engineer at Google DeepMind and a doctoral researcher at University College London, supervised by John Shawe-Taylor. I work on building scalable AI systems and multimodal models that can actively learn, understand, and interact with the real world.

In a past life, I was deep in the world of robotics. I earned a Bachelor’s degree in Electrical Engineering from the University of Sarajevo, followed by a Master’s in Robotics, Systems and Control at ETH Zurich. My thesis, completed externally at the Max Planck Institute for Intelligent Systems, focused on learning to grasp using tactile sensors.

Along the way, I worked as a teaching and research assistant and spent time at Rapyuta Robotics, where I developed navigation algorithms and worked on autonomous warehouse robots and end-to-end drone autonomy.

I’m passionate about strengthening the scientific community in Bosnia and Herzegovina and the broader region. I co-founded the Association for the Advancement of Science and Technology (ANNT) to help connect and mobilize Bosnian talent worldwide.

A lot of what I do on the side ties back to education — both organizing programs and building tools and material that make machine learning easier to pick up. Last year I co-organized EEML 2025 in Sarajevo, the prestigious Eastern European Machine Learning Summer School. It ended up being the largest in-person edition so far, with around 350 participants from about 50 countries (the wrap-up post has the full story). On the tools side, Blox is a small functional neural network library in JAX that I use for teaching and for my own experiments, and Möbius is a self-hosted personalized assistant I’ve been using to test ideas about long-running agents that build and maintain their own tools.