Although we like it or not, injuries are inevitable. Each year, 50% of Americans over the age of 18 develop a musculoskeletal injury that lasts longer than 3 months. However, only 30% of patients that utilize physical therapy services attend the visits that their insurance company authorizes. For patients to make progress in physical therapy, it isn't just about their physical endurance. It also heavily depends on the healthy mental state. The current physical therapy is time consuming, expensive and inaccessible. Often, people don't recover from their injuries properly simply because they don't stick to their treatments. Spark makes physical therapy fun, engaging and accessible in their personal environments through an Artificial Intelligence machine that uses a combination of two intelligent projectors and a 3D infrared sensor to overlay visuals and exercises tailored to the user’s injury and progress. Using emerging technologies of machine learning and artificial intelligence taking over, we wanted to SPARK to become the perfect personal physical therapist that could help users perform their daily exercises correctly with Spark's voice coach, and the motivating visuals set out in the patient's room. The visuals can be projected accordingly to the user's space as it recognizes objects such as tables, or sofas in the room to use these objects as a part of an exercise, and not as something users would need to remove during the experience. The role of machine learning in SPARK is to collect information during and after each exercise or routine on pain levels, and feedback. The constant communication between SPARK and the user is through voice control, so the users do not become frustrated from going back and forth of another controller.