Recognizing the critical need for advanced risk management in the rapidly expanding construction sector, this study aims to analyze and optimize construction risk mitigation strategies at a prominent Indonesian housing developer, enhancing project time efficiency and reducing decision-making uncertainty. The research methodology employs a quantitative descriptive approach, utilizing Markov Decision Process to model project risk dynamics and Reinforcement Learning, specifically the Q-Learning algorithm, to determine optimal mitigation policies. Data collection involved direct observation, in-depth interviews with project management, and analysis of historical project documentation from a housing project. Research findings demonstrate that the Q-Learning model effectively identifies and recommends adaptive mitigation strategies for various risk levels, providing optimal actions that significantly reduce project delays. The implementation of these data-driven strategies resulted in a notable improvement in project time efficiency, reducing the average project duration. Reproducibility, convergence, and sensitivity tests further validate the model's reliability and robustness, confirming its capacity to provide consistent and stable recommendations under diverse conditions.