Cardiopulmonary resuscitation (CPR) is critical to the prevention of death from cardiac arrest. Smartwatches can help first responders perform CPR correctly by collecting real-time acceleration data to determine compression rate and depth. We present an approach for a smartwatch application that uses motion data from the inertial sensors and an evolutionary algorithm ((μ + λ)-ES) to compute compression depth and frequency and provide corrective feedback. The application was optimized for out-of-hospital CPR and evaluated in a pilot study using a CPR training manikin as a reference system and an existing smartphone app for comparison.