Large-scale educational systems demand scalable, interpretable, and curriculum-aligned assessment models. Traditional approaches often fail to support high-frequency academic cycles or provide actionable insights. This paper introduces GAp (Grau de Aprendizagem), a semantic sensor designed to measure and estimate learning proficiency based on logistic transformations of performance deviations. Developed within a conceptual Digital Twin architecture, GAp bridges the gap between classical psychometrics and data-driven academic decision-making by operationalizing a probabilistic measurement layer. Two models are proposed: Model A retains a logistic formulation aligned with Item Response Theory (IRT), ensuring traceability to psychometric standards, while Model B offers a simplified version suitable for intuitive interpretation by non-specialist users. From a metrological perspective, IRT serves as the primary reference standard, Model A as a secondary reference, and Model B as a tertiary approximation. This hierarchical traceability enables flexible calibration based on institutional context and sustains the role of GAp as a semantic sensor within the educational data ecosystem. GAp depends only on actual (AV) and expected (NE) assessment scores, allowing for scalable deployment. A pilot test, conducted during the second semester of 2024, with over 70,000 students demonstrated strong concordance with IRT-derived rankings and minimal misclassification rates. Additionally, derived metrics—including measurements of learning gain (ΔGAp), discipline risk (RN, via negative semivariance), and preliminary signals for future Value-Added analyses—support institutional diagnostics and policy evaluation. In prospective use, GAp could feed optimization systems, such as those used by the Planning and Academic Efficiency Directorate (PEA), enriching scenario-based planning with cognitive insight. GAp time series have also shown potential for forecasting academic states using Markov chains, enabling predictive simulations within the Digital Twin framework. Within this architecture, GAp is positioned not merely as a metric, but as a cognitive signal driving feedback loops in a cyber-physical learning system.