Felix Leibfried

Contact me at felix.leibfried@gmail.com

Since May 2026, I have been an AI Engineer at Millennium in London, working on agentic AI for fixed income and commodities.

Before that, I was a Quantitative Developer at iSAM Funds, where I bridged the gap between research and execution by developing and productionising systematic equity trading strategies. Back then, my work focused on leveraging large language models (LLMs) to extract alpha from unstructured text data.

Earlier, I spent over three years as a Quantitative Researcher at Eisler Capital, where I designed machine-learning-driven systematic trading strategies for equity index and government bond futures, and developed a point-in-time machine learning pipelining framework for automated model training and evaluation within the firm's backtesting infrastructure.

Prior to entering the hedge fund space, I led a machine learning research team at Secondmind, an autonomous decision-making startup. There, I spearheaded the development of Bellman, an open-source library for model-based reinforcement learning designed to support both basic research and serve as a foundation for industrial applications, including the Secondmind spin-off Solvo.ai. I was also involved in the development of GPflux, a library for Bayesian deep learning and deep Gaussian processes aimed at providing flexible probabilistic models for approximate inference.

Before becoming a team lead, my research at Secondmind focused on reinforcement learning, with a specific emphasis on information-theoretically motivated algorithms, alongside contributions to Bayesian deep learning. On the applied side, I worked on several client-related projects, including: developing a model-based reinforcement learning agent for systematic trading—which moved from live deployment in the Secondmind fund to the strategic spin-off Gradient Systems; creating a framework for tuneable artificial intelligence for a large-scale digital gaming environment; and leading efforts to develop customised Gaussian process solutions for the recommendation system of a multinational educational publishing corporation's e-learning platform.

Prior to Secondmind, I interned at Microsoft Research, where I devised the first method to jointly predict video frames and game scores in Atari games from past video frames and agent actions using deep convolutional networks—solving a then-unresolved research problem.

I completed my PhD in artificial intelligence at the Max Planck Institutes for Intelligent Systems and Biological Cybernetics in Tübingen, Germany. My educational background is in computer science with a focus on artificial intelligence (equiv. BSc+MSc), and I also hold a degree in biology with a focus on neuroscience (equiv. BSc+MSc), both from the Julius Maximilian University of Würzburg, Germany.

See Google Scholar for my publications.

Page updated

Google Sites

Report abuse