This research-to-practice paper details implementing an educational approach designed to equip undergraduate students with the skills to develop healthcare applications and contribute to this domain using Medical AI by bridging the gap between the AI and the medical domain. This gap includes three challenges: medical data acquisition, strict privacy regulations in handling medical data, and a disconnect between the medical domain's requirements and AI's capabilities. To handle these challenges, We structured our educational approach into three phases: The first one is the educational AI stack which includes five learning stages, the second one is the medical foundation stack, which includes two working stages and the third one is system development, deployment, operation and testing. We implemented this approach within three Senior Design Projects focused on diagnosing skin cancer, breast cancer predication, and assessing allergy risks based on personal and biological data. The approach begins with a foundational AI stack, including Tensor-Flow, Scikit-Learn, and Keras, emphasizing transfer learning and model architecture selection customization. The second phase included the medical data acquisition process, NDA (Non-disclosure agreement) signing by the participants in these projects, and data handling, preprocessing, model selection, and training. The final phase included systems development, Deployment, and operation and testing. Students achieved significant results, including a 94% accuracy rate in skin cancer detection and over 85% precision in breast cancer prediction, including the tumor grade, stage, recurrence, and survival in the second project. And an 88% accuracy in allergy prediction using biological information, including Skin color and skin condition. They integrated the trained models with three developed web applications for the end users. Following the integration process, they deployed these applications into AWS and GCP. They compared the trained models between these two cloud platforms to determine their AI models' most effective deployment environment considering the model accuracy and cost. As a final step, the students tested the deployed applications within five stages, including the Go-live test, system performance, user satisfaction, model accuracy, and security status. Our approach aligns with ABET accreditation standards, focusing on the practical application of medical AI.

This research-to-practice full paper represents our work to comprehensively analyze cloud computing benchmarking with integrated AI-driven healthcare projects. As time progresses, the need to migrate a variety of system applications to the cloud has grown significantly. Yet, choosing a particular cloud service provider can be troublesome and proposes several challenges, whether computational or financial, depending on the domain of the application. Thus, the reliance on cloud platform benchmarking has become an essential skill engineers must develop in order to make informed and efficient decisions. Healthcare applications, in particular, constitute an excellent from-ground- to-cloud example that we needed to analyze and study, especially from the perspective of Artificial Intelligence (AI) applications. At the moment, there is high demand for AI applications in the domain of healthcare seeking higher performance while lowering the costs of the services. Nonetheless, the healthcare domain is considered a tough field to experience and experiment to new graduates and proposes challenges to successfully cloudify the designated system applications. Therefore, we developed an effectual pedagogical approach capable of addressing those stacked challenges and preparing a new generation of competent engineers. Our work is designed to implement multiple stages sequentially and individually. The first stage commences by introducing students to various concepts as a foundation for their core work, covering Machine Learning (ML) basics, healthcare domain knowledge basics, finding a solution via ground and cloud computation, and benchmarking. Students then attempted to address a chosen healthcare problem using AI techniques. Finally, they solve the proposed healthcare challenge and successfully benchmark the cloud platforms of both Google Cloud Platform (GCP) and Amazon Web Services (AWS). Students, as instructed, relied on concepts of metrics, measurements, criteria, and their disparities to accurately assess each cloud. This paper efficiently documents and analyzes the results of each proposed stage based on the represented students' deliverables and the supervisors' feedback to reflect the effectiveness of the suggested teaching approach.

The domain of software design is gaining a long overdue recognition as a vital discipline within software engineering, necessitating innovative approaches to its teaching in undergraduate education. Despite the growing importance of software design, academic programs often treat it as a secondary skill, overshadowed by the strong emphasis on coding. Only a few schools offer a design degree or dedicated design courses. This disparity between industry demands and educational practices underscores the need for novel pedagogical strategies. In our innovative work, we discuss a new way of effectively teaching software design-by-analogy for undergraduates to help them rapidly acquire the essential skills needed to design complex software without getting entangled in complex code generation and management. Software design does not necessarily follow the same clear delineation/separation between modules and components naturally apparent in tangible engineering domains. We employ “Conceptual Blending” to help students map their everyday experiences onto software design concepts. The process begins with students analyzing a simple two-arm watch to identify its user interface and create a finite state automaton for its interaction design. Success rates decline as the complexity of the watches increases, underscoring the software design challenges. By comparing these exercises to software interfaces, students learn to apply design techniques such as navigation modeling and prototyping, ensuring they can create intuitive, user-friendly software.

The current approach in software engineering curricula largely stresses learning programming principles, code building, and large-scale testing but often neglects the importance of applying domain-specific knowledge and original ideas in creating practical software solutions. To tackle this shortcoming, we orchestrated an experiment encouraging students to comprehend the necessity of domain expertise and initial high-level design before plunging into coding and system development. In our experiment, students were tasked with designing a software application that calculates the maximum cube volume from a given length of wood. The first phase, which saw the students mainly focus on coding, resulted in a flawed calculation as it disregarded two dimensions of the wood. The second phase of the experiment introduced a physical component: students had to physically build the cube using their software. The realization of their calculation error prompted an understanding of the importance of careful design and prototyping. Afterward, students adjusted their approach, incorporating careful drawing and calculation into their process. This led to the correct algorithms to find the cube's maximum volume and precise dimensions. Further projects involving different geometric shapes reinforced this learning. The experiment demonstrated the value of incorporating domain knowledge and user needs at the onset of the software design process, proving the effectiveness of a more physically engaged, project-based learning approach.

The Robotics Innovative Technologies and Education (RITE) Lab provides an interdisciplinary environment for designing and constructing intelligent social robots leveraging artificial intelligence (AI) and Human-Computer Interaction (HCI). This paper is divided into two parts. In the first part, we delineate five distinct generations of robotics technology identified through a comprehensive literature review. The second part showcases our lab's successful strategy in covering these five generations by offering a comprehensive multi-disciplinary theoretical and hands-on experience using nine individual modules. This is achieved by using a unique ‘black box/white box’ approach and continuous improvement since 2012, ensuring a 100% success rate, with students capable of building and programming their robots from scratch. Our program's broad scope spans imminent robotics innovations (2023-2028) and envisions long-term future developments. This paper serves as a blueprint for similar educational endeavors, encapsulating the RITE Lab's successful ten-year journey.

Team-Based Learning (TBL), a pedagogical approach that positively influences classroom attendance, still needs help with student absenteeism. Current attendance tracking tools are designed for something other than TBL environments and require manual interaction from the instructor or students, consuming valuable class time. This paper introduces a novel approach to this problem, proposing an automated attendance tool for TBL classes using fixed sensors, either a camera or a Light Detection and Ranging (LiDAR), combined with a machine learning classification algorithm. The paper delves into a comparative study of using cameras and LiDAR for this purpose, evaluating them based on privacy, accuracy, efficiency, and perceptions of students and instructors. The results indicate that while both methods successfully record attendance, the LiDAR system was more efficient and reliable. Although the camera offered a higher accuracy rate, a better customized LiDAR dataset designed for the classroom environment could enhance the machine learning algorithm's accuracy in identifying students and recording attendance. Finally, the LiDAR system was favored by both students and instructors for its ease of use, non-intrusiveness, and privacy preservation.

Driver distraction is one of the primary reasons for fatal car accidents. Modern cars with advanced infotainment systems often take some cognitive attention away from the road, consequently causing more distraction. Driver behavior analysis can be used to address the driver distraction problem. Three important features of intelligence and cognition are perception, attention and sensory memory. In this work, we use a stacked LSTM network with attention to detect driver distraction using driving data and compare this model with both stacked LSTM and MLP models to show the positive effect of using attention mechanism on the model's performance. We conducted an experiment with eight driving scenarios and collected a large dataset of driving data. First, an MLP was built to detect driver distraction. Next, we increased the intelligence level of the system by using an LSTM network. Third, we used the attention mechanism increment on the top of the LSTM model to enhance the model performance. We show that these three increments increase intelligence by reducing train and test error. The minimum train and test error of the stacked LSTM were 0.57 and 0.9 that were 0.4 less than the MLP minimum train and test error. Adding attention to the stacked LSTM model decreased the train and test error to 0.69 and 0.75. Results also show diminished the overfitting problem and reduction in computational expenses when adding attention.

Driver distraction is one of the primary causes of fatal car accidents in U.S. Analyzing driver behavior using different types of data including driving data, driver status or a combination of them is an emerging machine learning solution to detect the distraction level and notify the driver. Deep learning methods such as recurrent neural networks outperform other machine learning methods in car safety applications. In this paper, we used time-sequenced driving data that we collected in eight driving contexts to measure the driver distraction level. Our RNN is also capable of detecting the type of behavior that caused distraction. We used the driver interaction with the car infotainment system as the distracting activity. Two types of LSTM networks were used including bidirectional LSTM network and attention network. We compare the performance of these two complex networks to that of the simple LSTM in estimating driver behavior. We show that our attention network outperforms the other two, while adding bidirectional LSTM networks enhanced the training process of simple LSTM network.

An information technology (IT) tool that can guide STEM educators through the complex task of course design development, ensure tight alignment between various components of an instructional module, and provide relevant information about research-based pedagogical and assessment strategies will be of great value. A team of researchers is engaged in a User-Centered Design (UCD) approach to develop the Instructional Module Development System (IMODS), i.e., a software program that facilitates course design. In this paper the authors present the high-level design of the IMODS and demonstrate its use in the development of the curriculum for an introductory software engineering course.