The proliferation of internet of things solutions has been gaining importance, which are now widely adopted in residential and commercial buildings. These solutions can benefit from the adoption of artificial intelligence models to create intelligent solutions and provide supported actions and decisions to assist users and building operators. The evolution of technology, hardware, and software enabled the capability of having artificial intelligence models deployed in the edge layer and on the internet of things’ devices. However, further studies to test these capabilities are needed to validate the feasibility of having artificial intelligence-based models near the devices, resources, and users. The proposed solution presented in this paper will assess the feasibility of having in the same microcontroller three machine learning models while being able to read and measure multiple signals from sensors and communicate the sensor’s data and the models’ outputs to a streaming communication protocol. In this work, two neural networks will be used for temperature forecast and CO2 forecast, and a random forest model will be used to classify, with true or false, the occupancy of a room. All the data will be communicated using the message queuing telemetry transport protocol. The results seem very promising, and the microcontroller was able to perform the given tasks.