Catch up on the latest talks from Munich Datageeks — fresh insights, cutting-edge tech, and real-world data stories from top minds in the field. Stay curious and stay inspired!
Learn more about the hottest topic's in data and AI.
ML project failures—from misaligned goals to ignored baselines—stem from one root cause: lack of design documentation. Six real-world cases demonstrate how two weeks of upfront doc writing prevents months of wasted effort and embarrassing failures.
TNG's innovation team programmed a Unitree G1 humanoid robot to autonomously drive go-karts using creative hardware solutions and classical computer vision, while exploring reinforcement learning and foundation models for generalized robot intelligence.
Process mining case study at Munich's Großhadern hospital reveals that simple arithmetic mean outperforms AI in predicting surgery durations. Domain expertise proves essential for meaningful analysis, and augmentation beats automation.
Sampling-based Bayesian deep learning outperforms traditional approximations in both uncertainty quantification and predictive accuracy. Modern algorithms like microcanonical Langevin Monte Carlo enable efficient sampling of networks with millions of parameters.
Framework for evaluating LLMs in production classification systems using three components: golden datasets for ground truth benchmarking, LLMs as judge for semantic validation, and human feedback for real-world performance tracking.
