About

I’m a principal scientist at Symbolica, building neural networks that process structured data based on the Categorical Deep Learning programme.

I defended my PhD thesis “Fundamental Components of Deep Learning: A category-theoretic approach” in the Mathematically Structured Programming Group at the University of Strathclyde in Glasgow. My advisor was Neil Ghani. During my PhD, I have written a number of papers exploring the intersection of category theory, machine learning, game theory, and cybernetics.

I previously studied computer science at the Faculty of Electrical Engineering and Computing (FER) in Zagreb, Croatia. I was a part of TakeLab. I’ve done quite a bit of programming work there, reimplementing a number of neural network architectures such as Differentiable Neural Computer, Generative Adversarial Networks, Synthetic Gradients and made even my own automatic differentiation framework. Attempting to make my thinking more abstract and precise, I discovered category theory, which greatly shaped how I think about research. My studies in Zagreb culminated with me writing my master thesis on the topic of formalization of Cycle-Consistent Generative Adversarial Neural Networks (CycleGAN) in the language of category theory. This thesis was turned into a paper Learning Functors using Gradient Descent and accepted at the ACT conference.

My CV is here.

Contact

Click to reveal email. You can also find me on Twitter and Github.

You can subscribe to my RSS feed.