In recent years, the severity of forest fires has reached worrying levels both internationally and nationally. However, thanks to the advance of technology, it is possible to predict forest fires occurrence and magnitude through Machine Learning models specially developed for this purpose. To achieve this goal, this paper describes the development of an automated data pipeline in the Python programming language that generates a forest. fires dataset specific to Pinamar area, thus allowing the subsequent training of predictive fire models. It is also configurable to gather meteorological, topographical and fuel data from other geographical areas.