The Internet of Things (IoT) has experienced rapid expansion as a result of the widespread adoption of smart devices, sensors, and embedded systems capable of collecting and processing large volumes of data. The traditional reliance on cloud computing for data processing has led to challenges related to response time, bandwidth consumption, privacy, and service continuity. In this context, the concept of Edge Artificial Intelligence (Edge AI) has emerged as an approach that combines artificial intelligence technologies with edge computing to move part of the data analysis and decision-making processes to devices or nodes located close to the data source. This article aims to review the concept of Edge AI, its fundamental architecture, and its major applications in IoT systems, with particular emphasis on its role in reducing latency, improving resource utilization efficiency, and enhancing privacy and cybersecurity. The article also discusses the major challenges associated with deploying AI models on edge devices, including limited computational capabilities, memory and energy constraints, difficulties in updating models, device and protocol heterogeneity, as well as security challenges. Recent studies indicate that the integration of artificial intelligence with edge computing represents an important direction in computer engineering, particularly for applications requiring real-time data processing and decision-making. Keywords: Computer Engineering, Artificial Intelligence, Edge AI, Internet of Things, Edge Computing, Machine Learning, Cybersecurity, Embedded Systems. 1. Introduction The rapid development of computing and communication technologies has led to the widespread deployment of network-connected devices, ranging from smartphones and wearable devices to industrial sensors, smart city systems, and autonomous vehicles. These devices generate large volumes of data that require processing and analysis to extract meaningful information and support appropriate decision-making. Many IoT systems initially relied on transmitting data to cloud servers for processing and analysis. Despite the advantages provided by cloud computing, transmitting all data to data centers may increase latency and bandwidth consumption and may also raise privacy concerns when sensitive data are involved. Edge computing has emerged as a solution to address these issues by performing data processing closer to the data source. With the advancement of artificial intelligence technologies, it has become possible to deploy machine learning and deep learning models on edge devices, a paradigm known as Edge AI. Scientific reviews indicate that this approach aims to achieve faster data processing while reducing reliance on cloud infrastructure. 2. Concept of Edge Artificial Intelligence Edge Artificial Intelligence can be defined as the integration of artificial intelligence algorithms and techniques with edge computing infrastructure, whereby processes such as inference and data analysis are performed close to where the data are generated. This model differs from traditional cloud-based artificial intelligence, which relies heavily on transmitting data to centralized servers. In an Edge AI model, a smart camera, sensor, smartphone, or embedded controller can perform part of the analysis locally. Studies indicate that Edge AI can provide benefits such as reduced latency, reduced amounts of data transmitted over the network, improved privacy, and support for applications requiring near-instantaneous decision-making. 3. General Architecture of Edge AI Systems The system architecture can be conceptualized as consisting of three main layers: 3.1 Device and Sensor Layer This layer includes sensors, cameras, wearable devices, and embedded systems that collect data from the surrounding environment. 3.2 Edge Computing Layer This layer represents the area where data are processed close to their source. It may include edge servers, smart gateways, or embedded computers. 3.3 Cloud Computing Layer The cloud layer is used to perform operations requiring substantial computational resources, such as training large models, storing massive volumes of data, and managing the system on a large scale. Accordingly, these layers can operate in an integrated manner, with rapid inference processes performed at the edge, while the cloud is utilized for operations requiring greater computational resources. 4. Applications of Edge Artificial Intelligence 4.1 Smart Healthcare Edge AI can be used in wearable devices and medical monitoring systems to analyze physiological data close to the patient. This enables certain abnormal patterns to be detected more rapidly and reduces the need to continuously transmit raw data to servers. This approach can also be particularly important for applications that require the preservation of health-data privacy. 4.2 Smart Cities Edge AI can be integrated with smart city cameras and sensors to analyze traffic patterns, detect congestion, and manage traffic signals with greater responsiveness. 4.3 Smart Industry In Industrial IoT environments, Edge AI can be used to analyze data generated by machinery and equipment and detect abnormal patterns, thereby supporting the development of predictive maintenance systems. Research in industrial Edge AI has focused on achieving rapid response, customization, and privacy preservation as key challenges in Industrial IoT applications. 4.4 Vehicles and Autonomous Systems Smart vehicles need to analyze data from cameras, radars, and sensors within extremely short time intervals. Therefore, executing certain AI algorithms close to the vehicle can reduce reliance on continuous connectivity to the cloud. 4.5 Cybersecurity Cybersecurity is an important application area for Edge AI, as machine learning algorithms can be used to detect abnormal behaviors and intrusion attempts within IoT networks. Recent reviews indicate growing use of artificial intelligence and machine learning in intrusion detection systems and anomaly analysis within IoT environments, while challenges related to resources, energy, communications, and model deployment on edge devices continue to exist. 5. Role of Edge AI in Cybersecurity IoT systems are characterized by significant diversity in devices, systems, and communication protocols. Moreover, some IoT devices have limited resources, making the security of these environments a complex engineering challenge. Artificial intelligence can be used for: Detecting cyberattacks. Detecting abnormal behavior. Classifying different types of attacks. Analyzing network traffic. Detecting malware. Supporting Intrusion Detection Systems (IDS). Improving incident response. Recent studies demonstrate that integrating AI with IoT-edge systems can support intrusion and anomaly detection. However, model evaluation continues to face challenges related to variations in datasets, memory and energy requirements, response time, and reproducibility of results. 6. Engineering Challenges Despite the advantages offered by Edge AI, its practical implementation faces several challenges. 6.1 Resource Limitations Many edge devices have limited capabilities in terms of: Memory. Processing power. Energy. Storage. Therefore, large AI models cannot always be executed directly on such devices. 6.2 Energy Consumption Energy consumption is an important factor, particularly for battery-powered devices such as sensors and wearable devices. 6.3 AI Model Size Deep learning models can be large and complex, making their deployment on resource-constrained devices challenging. Several approaches have been proposed to address this issue, including Model Compression, Quantization, and Knowledge Distillation, with the aim of reducing model requirements while maintaining an acceptable level of performance. 6.4 Security and Privacy Although local data processing can reduce the need to transmit raw data to the cloud, edge devices themselves may be vulnerable to attacks. Therefore, appropriate security mechanisms must be designed to protect both data and models. 6.5 Scalability As the number of edge devices increases, managing models, updating them, and monitoring their performance become increasingly complex. 7. Federated Learning and Its Role in Edge AI Federated Learning represents one of the promising approaches for addressing certain privacy challenges in Edge AI systems. Instead of transmitting raw data from all devices to a central server, the model can be trained locally on individual devices, after which model updates or parameters are transmitted to a central server for aggregation. This approach reduces the need to transfer raw data, making it particularly suitable for certain IoT applications that handle sensitive information. However, Federated Learning introduces new challenges, including communication costs, device heterogeneity, variations in data across devices, and security threats that may target the training process itself. 8. Future Research Directions Recent literature highlights several directions that may constitute important research areas in computer engineering, including: Designing specialized processors for Edge AI. Developing small and lightweight artificial intelligence models. Improving the energy efficiency of edge devices. Integrating Edge AI with 5G and beyond-5G networks. Developing intelligent real-time intrusion detection systems. Using Federated Learning to protect data privacy. Developing Explainable AI for edge devices. Integrating Edge AI with self-protecting cyber systems. Developing standardized criteria for evaluating model performance on edge devices. Investigating the use of Large Language Models (LLMs) in IoT environments while reducing computational requirements. Recent research indicates growing interest in techniques such as Federated Learning, Explainable AI, and Deep Learning, alongside the need to improve model efficiency and reduce computational requirements in IoT environments. 9. Discussion The analysis of the literature demonstrates that the relationship between edge computing and artificial intelligence represents a multidisciplinary direction that brings together computer engineering, embedded systems, networking, artificial intelligence, and cybersecurity. The success of Edge AI systems is not limited to selecting an algorithm with high accuracy. It also depends on a set of engineering metrics, including latency, energy consumption, memory requirements, model size, bandwidth, security, and scalability. Recent studies indicate that one of the major challenges in this field is the transition from laboratory-based models to real-world systems. Some studies continue to focus on limited benchmark datasets, while metrics such as energy consumption, latency, and communication costs are not consistently reported. 10. Conclusion Edge Artificial Intelligence represents an important direction in the modern development of computer engineering due to its ability to combine artificial intelligence with computing capabilities located close to the data source. This approach contributes to reducing latency and minimizing the volume of data that needs to be transmitted to the cloud. It can also support privacy and efficiency across a wide range of IoT applications. Nevertheless, the successful implementation of Edge AI depends on addressing several challenges related to computational resources, energy consumption, security, scalability, and model management. Therefore, future research should not focus solely on improving the accuracy of artificial intelligence algorithms but should also address model efficiency, deployability, security, and performance in real-world environments. Accordingly, the integration of artificial intelligence, edge computing, the Internet of Things, and cybersecurity represents a promising research area for computer engineering students and researchers. It offers numerous opportunities for developing intelligent, secure, and efficient systems capable of operating in real time. References Gill, S. S., et al. (2024). Edge AI: A Taxonomy, Systematic Review and Future Directions. arXiv. Zhou, Z., Chen, X., Li, E., Zeng, L., Luo, K., & Zhang, J. (2019). Edge Intelligence: Paving the Last Mile of Artificial Intelligence With Edge Computing. Proceedings of the IEEE, 107(8), 1738–1762. DOI: 10.1109/JPROC.2019.2918951. Zhou, Z., et al. (2020). Edge Intelligence: The Confluence of Edge Computing and Artificial Intelligence. IEEE Internet of Things Journal, 7(8), 7457–7469. DOI: 10.1109/JIOT.2020.2984887. Mallidi, S. K. R., & Ramisetty, R. R. (2025). Advancements in training and deployment strategies for AI-based intrusion detection systems in IoT: A systematic literature review. Discover Internet of Things, 5, Article 8. DOI: 10.1007/s43926-025-00099-4. Alioanei, C., & Popescu, N. (2025). AI-Based Solutions for Security and Resource Optimization in IoT Environments: A Systematic Review. Information, 16(10), 841. DOI: 10.3390/info16100841. Xue, Q., Xue, P., Wang, Z., & Ma, H. (2026). Artificial Intelligence for Cybersecurity in IoT-Edge Systems: A Structured Review of Methods, Datasets, Evaluation, and Deployment Challenges. Electronics, 15(11), 2409. Edge Computing for Internet of Things IoT—A Review of Architectures, Challenges, and Emerging Trends. IEEE, 2025. DOI: 10.1109/ICRTEECT67512.2025.11448963. The Role of Large Language Models in IoT Security: A Systematic Review of Advances, Challenges, and Opportunities. Internet of Things, 34, 101735, 2025. DOI: 10.1016/j.iot.2025.101735. The Role of Large Language Models in Addressing IoT Challenges: A Systematic Literature Review. Future Generation Computer Systems, 171, 107829, 2025. DOI: 10.1016/j.future.2025.107829. Edge AI: A Comprehensive Survey of Technologies, Applications, and Challenges. IEEE, 2024. DOI: 10.1109/ACET61898.2024.10730112.