Bruno Gavranović

WORK HISTORY

EDUCATION

SKILLS

ABOUT BRUNO GAVRANOVIĆ

I\'m building neural networks that generate provably correct code, and the software infrastructure for training them.My work is driven by a long-term research programme [1] on Categorical Deep Learning [2]. I pursue this work through Coend [3], a company I founded dedicated to it. I am also affiliated with Glasgow Lab for AI Verification [4], where I work on structured tactics [5] for theorem proving.I hold a PhD in computer science, with a specialisation in category theory. My thesis \"Fundamental Components of Deep Learning: A category-theoretic approach\"[6] is work I am extremely proud of, go check it out!On the research side I work on three fronts:1. New architectures: Understanding old, and designing new neural networks that natively consume and produce structured data2. Mathematical Foundations: Building the mathematics necessary to state precisely what it means to generalise, especially on data structures recursive in nature3. Infrastructure: Building the stack required to train such networks in dependently-typed languages: tensor processing, automatic differentiation and elaborator integrationIn all of these, I use category theory, the mathematics of structure and composition, as a central glue.My most recent project is TensorType [7]: a framework for type-safe, pure functional and non-cubical tensor processing.[1]: https://categoricaldeeplearning.com/[3]: https://glaive-research.org/[5]: https://github.com/bgavran/TypeSafe_Tensors