Strategic intelligence and data science professional with more than eight years’ experience in security, policy and analytical work within the public sector. Specialises in applying machine learning, statistical analysis, stochastic optimisation and data-driven decision support to national, economic and maritime security challenges. Research interests include risk-aware artificial intelligence, human-AI decision support, anomaly detection, NLP-based deception analysis, maritime security and policy-ready analytics. Combines domain expertise in strategic intelligence with emerging technologies to translate complex data into actionable evidence.

From Prediction to Decision: A Hybrid Machine Learning and Stochastic Optimisation Framework for Risk-Aware Decision Support                                                                                   Abstract: Artificial intelligence systems are increasingly capable of generating accurate predictions, yet high-stakes decision environments require more than predictive accuracy. Decision-makers must determine how limited resources should be allocated when information is uncertain, risks vary across alternatives, and future outcomes cannot be known with certainty. This study presents a hybrid decision-support framework integrating machine learning with two-stage programming to bridge the gap between predictive and prescriptive analytics. The framework uses machine-learning models to estimate behavioral characteristics, reliability and risk from historical and simulated data. These predictions are subsequently incorporated into a two-stage stochastic optimisation model in which initial resource-allocation decisions as uncertain outcomes are observed, followed by recourse decisions as additional information becomes available. The approach is demonstrated using a synthetic intelligence source-management environment designed to represet decision-making under uncertainty without reliance on operational or classified data. The study evaluates predictive performance, resource allocation, expected cost, risk exposure and decision outcomes under alternative optimisation strategies. It focuses on integrating model predictions with operational constraints and human oversight, rather than treating predictive accuracy as the final objective. The findings demonstrate the potential of combining machine learning with stochastic optimisation to develop risk-aware decision-support systems. The framework contributes to the emerging transition from prediction-oriented AI toward prescriptive systems capable of supporting transparent, uncertainty-aware and resource-constrained decision-making in high-stakes environments.                                                                                                                                                                                                                                                                                                                               Keywords: Two-stage stochastic programming, resource constraints, machine learning, uncertainty, behavioural characteristics