"The cost of training machines is becoming a problem". This is the title of an article from The Economist published in June 2020 that highlights the staggeringly unappreciated the financial impact of AI [38]. However, these costs on are not limited to monetary concerns but also accrue a concerning environmental toll. One round of training for some of the most complex machine learning models can emit millions of kilograms of carbon dioxide due to the electricity consumed . With the growing popularity of ML and the digitization across all sectors, there should be a growing awareness of the potential impact of these technologies on the environment and their potential contribution to climate change. Then, their use could be rationalized and steps can be taken to responsibilize the offset of their impact. We hereby propose CUMULATOR: an open-source API to quantify and report the carbon footprint of machine learning methods. As a demonstration, we integrated this API within an ML based medical research platform called Alg-E, which will be used in a large scale medical research project in Tanzania and Rwanda. We use CUMULATOR to analyse the trade-off between accuracy and carbon footprint within Alg-E and extend it with simple visualisations. In parallel, we also propose a Carbon Statement Protocol to quantify and report the carbon footprint of individual work, which uses this project as a proof-of-concept. This protocol and CUMULATOR thus comprise a great set of tools to report the carbon footprint of a large-scale medical research trial and EPFL research projects in the future.