Andrei-Octavian Brabete

I build machine learning systems, the infrastructure that makes training and serving large models fast, adaptive and reliable.

I have spent most of my career at the intersection of systems and machine learning. As a visiting researcher in the Large Scale ML Systems Group at the University of Edinburgh, I co-authored ServerlessLLM (OSDI ‘24), a system for locality-enhanced serverless LLM inference and helped open source it. Before that, I spent four years in quantitative finance, working on model traning, inference and big data at scale. Along the way I also explored real-time 3D video reconstruction with Gaussian Splatting during my time at Entrepreneur First.

My research career started at Imperial College London, where I worked on KungFu (OSDI ‘20), a system that makes distributed Deep Learning training adaptive. My Master’s Thesis on accelerating deep learning training on distributed GPU clusters was recognized as a Distinguished Project. I graduated with First Class Honours and a Prize for Academic Excellence.

I currently work as an AI Engineer in Equities Technology at Millennium Capital in London, where I am empowering equity research with AI, with a focus on quantitative AI research.