Boston-based construction company Suffolk has been collaborating with Smartvid.io , a specialist in image recognition based in Massachusetts, on the development of an AI engine nicknamed Vinnie for predicting accidents for one year now. Vinnie has been trained using hundreds of thousands of images from the Procore construction project management software, as well as data and photos from construction projects over the past decade. It should now be able to monitor entire construction sites and to recognize and warn about dangerous situations, such as when workers are not wearing protective clothing or are too close to a dangerous machine part.

During the test phase, Vinnie was fed photographs of dangerous situations, predicting 20% of incidents with an accuracy rate of 80%. In addition to a reduction in the risk of injury to workers, Suffolk calculates that the safety incidents avoided thanks to the engine will save construction companies an average of US$1.4 to US$3.6 million per year. Suffolk and Smartvid.io have now teamed up with competitors to form the Predictive Analytics Strategic Council , to further improve technology through the sharing of data between the companies involved.