Buildings account for 37% of US carbon emissions, and roughly 15% of global energy consumption. Despite transformative advances in AI across industries, most buildings today still operate as they did 40 years ago. Building optimization research is ongoing but mostly confined to dedicated conferences such as BuildSys or ASHRAE, limiting ML community engagement. Despite many open research problems, the potential for ML-driven impact on climate change, in both the short and long term, is enormous, perhaps more so than any other optimization problem. This workshop will engage the ML community in tackling building optimization and fostering collaboration across ML, HVAC, dynamic systems, and smart buildings. Our confirmed speakers from the smart buildings community, the ML community, and its emerging intersection will lead discussions around benchmarking standards, methodologies, and scalable solutions. Reducing the carbon footprint of buildings provides a direct way for the ICML community to benefit humanity and offset the carbon cost of AI. To engage the community, we will host a $10,000 Kaggle competition, where competitors predict building dynamics on an unreleased, real-world dataset from multiple buildings. The contest winner(s) will be invited to give a talk and share their methods.