Lorenz Richter is a co-founder, chief technology officer, and managing director of dida, the Berlin machine learning firm, and a mathematician who still publishes research on diffusion models while leading its machine learning team (dida; dida).
Studying mathematics in Berlin
He trained in stochastics and numerics at the Free University of Berlin and holds a doctorate, and he moved into deep learning algorithms from that mathematical base (dida). By dida’s account he has spent about ten years solving practical data science problems alongside his interest in theory (dida).
Leading dida
He became a managing director of dida in June 2021 and has led the company with Markus Düttmann since March 2025 (North Data). His view of the market is that ready-made AI tools fit only part of the work companies want to automate: “This is why we focus on tailoring AI solutions to specific use cases, with a focus on deep technical details and robust deployment” (dida).
Presenting research in Seoul, Tokyo, and Taipei
In 2026 he presented two papers on diffusion-based sampling at ICML in Seoul, one on a bridge matching sampler and one on tensor train diffusion for high-dimensional problems (dida). In July he spoke at DO x ML 2026 in Tokyo and ICOML 2026 in Taipei on the duality between sampling and optimization, building on work with Julius Berner and Karen Ullrich that links stochastic optimal control to diffusion-based generative modeling (dida).