ai systems

This study addresses the prevailing gap in AI education, which emphasizes model development while neglecting system engineering practices, leaving students ill-equipped to handle real-world challenges such as architectural design, deployment, and monitoring. To bridge this gap, the authors implemented a master’s-level course in which students built a movie recommendation system under realistic constraints, with a focus on integrating AI components into robust software systems, adopting data-driven machine learning practices, and cultivating systems-level thinking. Using a mixed-methods approach—combining analysis of student project artifacts with survey data—the research evaluates learners’ performance in architectural decision-making, integration of heterogeneous models, and adaptation to evolving requirements. Findings reveal common difficulties students encounter in AI system engineering and demonstrate the course’s effectiveness in addressing critical deficiencies in AI engineering education and enhancing systems-aware competencies.

This work addresses the challenge that existing AI systems struggle to dynamically observe, intervene in, and optimize agent behavior at runtime, making it difficult to simultaneously achieve high task success rates, low latency, token efficiency, reliability, and safety. To overcome this limitation, the paper proposes a novel runtime infrastructure layer situated between the model and the application, which treats the AI execution process itself as an optimizable object—departing from conventional approaches that restrict optimization to static model or log-level adjustments. This layer enables proactive intervention and multi-dimensional performance co-optimization through mechanisms such as runtime monitoring, real-time inference, adaptive memory management, fault recovery, and policy enforcement. Experimental results demonstrate that the proposed approach significantly enhances the holistic performance of long-horizon agent workflows across task success rate, response latency, token efficiency, system reliability, and safety compliance.

Traditional AI systems rely on fixed monolithic models, which struggle to dynamically allocate resources, decompose tasks, or update knowledge in response to varying inputs, leading to degraded performance and increased costs. This work proposes the first system-level design methodology for distributed composite AI systems, formulating a design space through workflow topologies and configuration choices and identifying eight core design patterns. The framework jointly optimizes model selection and runtime parameters, enabling task decomposition, multi-model orchestration, and explicit control logic, thereby facilitating a shift from static monolithic architectures toward dynamic, composable, and adaptive ones. Evaluated across three case studies, the approach reduces latency by up to 60% and cost by up to 71%, with only a 2.5–4 percentage point drop in accuracy.

This study addresses the widespread yet often insecure integration of AI components into software systems, which frequently overlooks critical security risks and can lead to malicious behaviors or data breaches. Through semi-structured interviews with 22 industry practitioners, the work systematically uncovers a pervasive neglect of security considerations during AI component selection and integration, revealing that functional performance overwhelmingly dominates decision-making while security is rarely evaluated. Drawing on established practices from traditional software supply chain security, the paper adapts and extends these principles to the AI context, proposing a set of lifecycle-spanning security-by-design guidelines. It further offers actionable recommendations tailored for developers, model providers, and researchers to foster more secure AI integration practices.

This study addresses the pervasive lack of verifiability, version control, observability, and traceability across the end-to-end AI pipeline—from data acquisition to inference. It proposes the first systematic modeling of the AI software supply chain as a four-layer architecture encompassing data ingestion, model training, model inference, and underlying infrastructure. Through this framework, the work identifies four structural security gaps overlooked by conventional mechanisms: behavioral coupling, insufficient rollback capability, absence of change awareness, and difficulties in lineage tracking. Leveraging software supply chain analysis, dependency resolution, and large-scale empirical measurement, the authors evaluate 48 production-grade open-source projects—comprising 4,664 direct and 11,508 transitive dependencies, totaling approximately 392 million lines of code—thereby quantifying the complexity and scale of latent risks inherent in modern AI supply chains.

Current AI research often treats models as static artifacts, overlooking the fundamental influence of training dynamics on critical properties such as capability, bias, robustness, and safety. This work proposes shifting the focus toward the training process itself to establish a science of AI centered on training dynamics. By analyzing the interactions among data, objectives, architectures, and optimizers, the paper develops a theoretical framework that is predictive, intervenable, and design-oriented. Integrating approaches from mechanistic interpretability, fairness, memory mechanisms, and simplicity biases, it uncovers causal links between early-training signals and final model behavior. The study systematically outlines key challenges and open problems, offering both theoretical pathways and practical foundations for extending scaling laws beyond performance to encompass multidimensional model attributes.

This study addresses the lack of systematic synthesis at the intersection of artificial intelligence (AI) and modeling and simulation (M&S) by proposing, for the first time, a structured framework based on the full M&S lifecycle—encompassing model construction, input modeling, execution, experimentation, validation, and output analysis. It elucidates the bidirectional integration mechanisms between AI and simulation: how AI enhances or substitutes traditional simulation components, and how simulation supports AI training and evaluation. Incorporating generative AI technologies such as large language models, the paper identifies representative application paradigms and integration approaches across each phase, synthesizes key achievements, and presents a conceptual roadmap tailored to the rapidly evolving ecosystem, while highlighting current limitations and open research challenges.

This work addresses the complexity and heavy reliance on expert knowledge in deploying edge AI models, particularly when adapting to hardware-specific inference runtimes such as Qualcomm QNN/SNPE. To tackle this challenge, the authors propose AIPC—the first AI agent–based automated deployment framework—that decomposes the deployment pipeline into standardized, verifiable stages. By integrating domain expertise through agent skills, auxiliary scripts, and iterative stage-wise validation loops, AIPC enables knowledge-guided verification and limited repair capabilities. Experimental results demonstrate that AIPC can complete end-to-end deployment of canonical vision models from PyTorch to QNN/SNPE within 7–20 minutes (at an API cost of approximately \$0.7–10 per run), effectively diagnose failures in complex models, and provide actionable repair suggestions, thereby substantially reducing the need for manual intervention.