Cerebral palsy (CP) is one of the most common physical disabilities of childhood, impacting over 10,000 children born each year in the United States. Caused by an injury near the time of birth, CP causes impaired movement and coordination that hinder activities of daily living. Most children with CP undergo multiple orthopedic and neurological surgeries to address neuromusculoskeletal impairments and improve movement. However, since each brain injury and movement pattern is unique, identifying the optimal treatments for each individual remains challenging and outcomes are highly variable. In this talk, I'll share our adventures and collaborations between engineers and surgeons at Gillette Children's Specialty Healthcare to quantify and improve movement for children with CP. We use machine learning and causal analyses to clarify (often unwritten) assumptions, test hypothesized mechanisms contributing to treatment outcomes, and develop patient-specific models to guide surgery and rehabilitation interventions. Specifically, I'll share three examples for how we have used multi-modal data from clinical gait analysis to evaluate (1) whether selective dorsal rhizotomy (a highly invasive procedure that cuts the spinal cord) improves walking efficiency, (2) whether motor control predicts outcomes after orthopedic surgery, and (3) whether personalized step-by-step models of treadmill training can be used to guide rehabilitation. We use Directed Acyclic Graphs to clarify clinician assumptions, identify critical adjustment sets, and develop interpretable models with Bayesian Additive Regression Trees - but we are ALWAYS open to your suggestions and improved methods to support mobility for children with CP and other developmental disabilities.