This presentation introduces, first, in a separate way, some issues and results we obtained during the past decade in Scheduling, Machine Learning, and Architecture, then, we explain how the context of Smart Buildings offers opportunities to glue main results to provide with a middleware able to manage the buildings. The key objective is to define and to implement the Operating System of the buildings, able to compute, to learn, to analyse, to monitor and to store, inside the buildings, all the data produced in the buildings, for privacy issues. This ongoing research is conducting with Qarnot Computing, a French SME, that initially was selling computing time slots on digital heaters at home, and which is now working on crypto heaters, digital irrigation for Smart Buildings. The concern of the presentation is thu, energy efficient computing, learning, analysis… as we had, two decades ago, with Desktop Grid Computing. This presentation was made possible thanks to the support of the French Embassy in Tokyo and corresponds to a joint effort with Mustapha Lebbah (Paris 13), Stephane Zuckerman (University of Cergy Pontoise), Yanik Ngoko (Qarnot), Tarek Menouer (now working with Umanis).