Despite major advances in artificial intelligence over the past decade, there are substantial challenges preventing current methods from impacting cancer research and care. For example, there may be limited data from which to learn (especially true in rare cancers such as multiple myeloma); there is often significant heterogeneity in patients’ disease stage and presentation at diagnosis; treatment strategies can vary widely; and symptoms and complications are not always recorded. To address these challenges, Dr. David Sontag and his team at the Massachusetts Institute of Technology (MIT) are developing machine learning and artificial intelligence algorithms that can learn models of cancer progression and treatment effectiveness directly from observational data found in disease registries and electronic medical records. The group will be able to capitalize on data from thousands of patients followed over many years, accessing significantly richer data than are typically available in any one randomized clinical trial. New methods developed will result in personalized treatment paradigms, personalized predictions of prognosis, improved clinical trial designs, earlier diagnosis of the cancer, its progression and complications, and new possibilities for drug targets. Software will be released free and open source under an MIT License to fully reproduce results with the twin goals of enabling other machine learning researchers to improve the methods and facilitating cancer researchers to adapt the methods to new settings. David and his research group believe that this work will open the door to machine learning methods being applied much more widely to cancer datasets.