š Hello, Iām Cormac
I am Cormac Kinsella (or Cinnsealach in Irish), a PhD candidate in mathematics at the Technical University of Munich, supervised by Prof. Massimo Fornasier at the Chair of Applied Numerical Analysis, and a member of the ATARI Lab, led by Prof. Majid Khadiv, at the Munich Institute of Robotics and Machine Intelligence. I hold a BSc in Mathematics from Trinity College Dublin (2023) and an MSc in Mathematics and Informatics from TUM (2026).
My research is on consensus-based optimisation (CBO) on manifolds: a swarm of interacting particles works together to find the global minimum of a non-convex function on a curved space. In robotics, I apply these methods to high-dimensional problems, in particular diverse scene generation. Given a scene, the goal is to find one example of every distinct way its objects can come to rest.
Before my PhD I worked as a software engineer, most recently at the BMW Group, where I built numerical libraries and data hubs. I am also an Irish speaker of the Cois Fharraige dialect of Connemara, and helped translate apps and games into Irish, including the official Irish translation of Among Us (covered by The Verge).
š Featured projects
- metal-linalg: GPU-accelerated batched QR, eigendecomposition and SVD on Apple Silicon, for MLX and PyTorch (details)
- holographic-tensor-networks: a Python library for holographic tensor networks on hyperbolic buildings (details)
- isomorphism: a backend-agnostic C++ tensor library for MLX, PyTorch and Eigen (details)
- riemannian-gaussian-sampler: exact sampling from the Riemannian Gaussian on SO(d) and Stiefel manifolds (details)
For my other projects, see my portfolio.
Feel free to reach out to me via email: cormac.c.00@gmail.com