The growing availability of data in semiconductor testing and digital healthcare creates new opportunities for data-driven optimization. At the same time, applications in these domains are highly performance-critical and must satisfy strict constraints. General-purpose optimization methods are rarely tailored to these settings, which often makes them computationally expensive and unsuitable for real-time operation. This thesis investigates data-driven optimization strategies that are explicitly trained for domain-specific requirements and designed for fast time to optimize using Deep Reinforcement Learning.