Accelerating ML development with optimized performance and cost