Automatic Radio Frequency signal classification is a challenging problem with multiple applications in signals intelligence and cognitive radio. In this talk we utilize a Python data science approach that include deep and shallow classifiers, AutoML, ensemble learning, expert feature extraction, importance analysis, and probability calibration. This work led to our team winning the Army Rapid Capability Office Signal Classification Challenge. We’ll discuss how each of these techniques went into generating the winning submission. The method developed is capable of correctly classifying signals at -10 dB SNR with over 60% accuracy and signals at +10 dB SNR with over 95% accuracy. Andres Vila Casado, Kyle Logue, Donna Branchevsky, Esteban Valles, Sebastian Olsen, Darren Semmen, Eugene Grayver and Alexander Utter