• Enable patient digital twin deployment via a knowledge-driven, modular framework • Provide clinical data access and simulation through a FHIR-compliant API • Publish MIMO ontology to standardize and increase transparency of AI model interfaces • Automate AI model integration using a manifest-based binding protocol • Benchmark stroke risk models with real-time, time-aware clinical data streams Background: The healthcare sector faces diverse challenges, including poor interoperability and a lack of personalized approaches, which limit patient outcomes. Ineffective data exchange and one-size-fits-all treatments fail to meet individual needs. Emerging technologies like digital twins (DTs), the semantic web, and AI show promise in tackling these obstacles. For this reason, we introduced CONNECTED, a conceptual multi-level framework that combines these techniques to deploy general-purpose patient DTs. Objective: This study assesses CONNECTED’s comprehensiveness, applicability, and utility for developing intelligent, personalized healthcare applications. Specifically, we deliver a preliminary version of the framework to predict future patient states and demonstrate its automation benefits in deploying semantically enriched, AI-powered patient DTs. Methods: We enhanced the CONNECTED architecture by providing a formal definition of DT and modularizing its core functionalities into microservices—Properties, State, Capabilities, and Manifest. The Manifest service facilitates AI model integration through the Model Interface Manifest Ontology (MIMO), enabling automatic data-to-model binding via a reasoner. Using the HeartBeatKG quality assessment tool, we validated MIMO and tested the internal logic by integrating a well-established stroke-risk model. Results: Our implementation comprehends: (1) deploying a FHIR-compliant, patient-centric API for clinical history access, real-time monitoring, and predictive simulation; (2) publishing MIMO; (3) establishing the Manifest protocol for seamless, general-purpose AI model integration tailored to individual patient profiles; and (4) a proof-of-concept benchmarking application comparing multiple stroke risk classifiers. Conclusion: CONNECTED establishes a flexible, scalable foundation for interoperable semantic patient DTs. Automation reduces technical overhead and enables users to focus on delivering personalized, insight-driven care.

Efficient job scheduling in high-performance computing (HPC) systems necessitates the simultaneous consideration of system-centric objectives, such as maximizing resource utilization, and user-centric objectives, such as minimizing job waiting times. In practice, the relative importance of these objectives is not static, but shifts dynamically in response to fluctuations in workload characteristics and system state. However, existing scheduling frameworks — including both traditional workload managers and reinforcement learning (RL)-based methods — typically rely on fixed policies or fixed reward functions that encode a predetermined combination of objectives. As a result, they lack the flexibility to adjust their scheduling priorities as workload intensity or system conditions change. In this work, we propose MetaPilot, a deep reinforcement learning-based controller designed to enable dynamic adaptation to shifting scheduling objectives. MetaPilot functions as an adaptive decision layer, continuously monitoring real-time system state and workload characteristics in order to select the most appropriate scheduling objective prior to each scheduling event. The selected objective is subsequently optimized by established scheduling frameworks, ensuring seamless integration with existing systems. Comprehensive experiments conducted on real-world HPC workload traces from both CPU-based (Physical) and GPU-enabled (Deeplearn) clusters demonstrate that MetaPilot achieves substantial improvements in both maximum and average job waiting times, as well as in resource utilization across CPU, GPU, and memory dimensions, relative to state-of-the-art baseline schedulers. Notably, MetaPilot reduces maximum waiting time by up to 19% on the Physical partition and achieves best-in-class utilization and balanced waiting times on the Deeplearn partition, thereby demonstrating robust adaptability and effectiveness across heterogeneous HPC workloads.