Equipping engineers for a world with AI is more than prompt engineering—many will design neural networks, foundation models and more to help solve problems in their careers.
Professionals across many sectors are experimenting with and integrating new generative AI tools into their work, but engineers have the potential to go deeper—not just designing prompts but tweaking the underlying algorithms.
Training future engineers to leverage this rapidly-growing technology has led to many Michigan Engineering professors to provide computer science training in disciplines outside computer science. Some scratch the surface by encouraging students to engage with tools based on large language models. Others introduce elements of machine learning in focused applications, and still others provide robust training in machine learning and foundation models within particular engineering contexts.
“Professional Issues and Design” (CEE 402) in civil and environmental engineering, is among those encouraging students to explore commercial AI tools. It assembles students from multiple disciplines into teams that then create a plan for a proposed land development project.
AI agents are welcome as thought partners and reviewers, but the students are expected to find citations for any information provided by the AI agent. And students are not allowed to include AI-generated content in the final deliverables. This course is jointly instructed by David Kelly, a lecturer in civil and environmental engineering; Jason McCormick, an Arthur F. Thurnau Professor in the same department; and Robert Sulewski, a teaching professor in technical communication.
Similarly, students in “Kinetics and Transport in Materials Engineering” (MATSCIE 335) are testing out ChatGPT as a coding partner, helping them to calculate, model and simulate problems that are far more advanced than they could ordinarily tackle using pencil-and-paper methods. Wenhao Sun, the Dow Early Career Professor of Engineering and an assistant professor of materials science and engineering, and Ashwin Shahani, an associate professor in the same department, introduced this aspect of the course.
Another example is “Socially Engaged Design of Nuclear Energy Technologies” (ENGR 100-910)—led by Aditi Verma, an assistant professor of nuclear engineering and radiological sciences, and Katie Snyder, a lecturer in technical communication—which uses AI image generation to help students visualize ideas for the exteriors of nuclear facilities. The ability to rapidly create a shared concept of what these designs could look like may facilitate the co-designing of future power plants with people in the communities where nuclear energy is sited.
Some graduate classes are exploring the use of machine learning, including large language models in their final projects—for instance in “Multidisciplinary Design Optimization (AE 588) and “Aerodynamic design (AE 740), taught by Joaquim Martins, the Pauline M. Sherman Collegiate Professor.
Beyond prompting existing AI agents, many courses are also giving students from an array of backgrounds a look under the hood in AI and machine learning, and senior-level courses often include graduate students.
As neural networks initially emerged from computer science, often built to serve the needs of robotics, it’s not surprising that these disciplines have some of the most robust offerings. Electrical Engineering and Computer Science (EECS) includes a long-running course series preparing students to develop neural networks, machine learning algorithms and AI, including foundation models like large language models. The two divisions that compose EECS, electrical and computer engineering and computer science and engineering, continue to develop new courses as the field evolves.
Expected to launch in winter 2026, the senior-level “Computer Engineering for and with AI/ML” (EECS 498) will teach students how AI and machine learning systems are used, how to design computer systems to support these applications, and how AI and machine learning can contribute to the design of those computer systems. The project-based course is being developed by Robert Dick, a professor of electrical and computer engineering.
Like students in EECS, roboticists need coding from the start. Robotics approaches that training by weaving the math into the applications from the beginning, starting with “Computational Linear Algebra” (ROB 101), which lays the groundwork for future AI and machine learning capabilities through hands-on programming of robots. The development of ROB 101 was led by Jessy Grizzle, the Elmer G. Gilbert Distinguished University Professor and Jerry W. and Carol L. Levin Professor of Engineering.
“Introduction to AI and Programming” (ROB 102) familiarizes students with programming autonomy and AI algorithms. It covers C++ and high-level scientific programming languages, autonomous navigation, search algorithms and entry-level neural networks. Chad Jenkins, a professor of robotics, led the development of ROB 102.
Bridging the senior level of undergraduate degrees and graduate coursework, “Deep Learning and Robot Perception” (ROB 498/599) teaches the fundamentals of how robots interpret sensor data through neural networks. Students first learn to program neural networks and then apply their skills by reproducing and implementing a state-of-the-art neural network that they find in the scientific literature. The development of this course was also led by Jenkins.
Mechanical Engineering is incorporating machine learning into the discipline with “Data Science for Manufacturing Quality Control” (ME 401), developed by Chenhui Shao, an associate professor in the department. It aims to teach skills related to data science and machine learning that are geared toward mechanical engineering students and can be used in manufacturing careers.
Machine learning is also finding a place in Nuclear Engineering and Radiological Sciences with “Nuclear Power Reactors” (NERS 442), which Majdi Radaideh, an assistant professor in the department, is evolving. He introduces students to neural networks for calculations related to nuclear reactor criticality. The course ends with a final project in which students develop deep neural networks for nuclear reactor power control, optimizing performance.
Radaideh is also updating “Nuclear Reactor Safety” (NERS 462). It lays the groundwork for machine-learning-intensive, graduate-level courses. With an emphasis on statistics and inference in the context of safety analysis, students learn to use AI and machine learning methods to quantify uncertainty in reactor simulations.
Outside conventional coursework, the Multidisciplinary Design Program enables cohorts of students to work with sponsors in industry and U-M units to solve real-world problems. Currently, 151 of the 800+ current MDP students are engaging in projects applying AI. Applications include modeling the human body for auto safety research, investigating spending habits for financial planning, tools to review resumes and grant applications, identifying improper medical billing, and autonomously guiding hospital stretchers. These experiences deepen technical mastery while cultivating the problem-solving and leadership skills that classroom instruction alone cannot fully develop, says program director Gail Hohner.
Many engineering departments are training students in AI and machine learning particularly for its utility in making sense of complex data. Rather than sending students to EECS for this information, offering these courses within different disciplines means students learn to use computer science tools in the context of the kinds of problems they will solve as professionals.
Climate and Space Sciences and Engineering, which relies heavily on data-informed physics models, provides a way for students to get started at the senior level with “Machine Learning for Earth and Environmental Sciences” (CLIMATE & SPACE 405). It starts with foundational mathematical tools and builds to various types of neural networks, always grounded in the topics of the discipline such as hydrology and climate science. It was developed by Mohammed Ombadi, an assistant professor in the department, intending to help students harness the power of big environmental data to gain insights into complex system behaviors.
Industrial and Operations Engineering has launched the course “Advanced Data Analytics” (IOE 473). Students who take this course can expect a deep understanding of modern tools and techniques in machine learning and AI, including the gathering and preparation of data ahead of employing machine learning methods to solve real-world challenges. The course, developed by assistant professor Salar Fattahi, also hosts guest lectures to expose students to the ways that machine learning is used in industry.
Civil and Environmental Engineering offers a series of three courses with expanded AI content, starting with “Computational Methods for Engineers and Scientists” (CEE 303), which introduces Python programming and computational thinking. Students may move on to “Statistical Methods for Data Analysis and Uncertainty Modeling” (CEE 373) and “Sensors and Data Acquisition” (CEE 375) to apply these skills in collecting and interpreting data. McCormick; Jeff Scruggs, a professor in the department; and Branko Kerkez, an Arthur F. Thurnau associate professor in the department, updated these courses.
“Machine Learning for Infrastructure Systems” (CEE 554) equips students with essential machine learning techniques to analyze complex data, enabling the development of smarter, more resilient infrastructure. The course integrates theoretical foundations with real-world case studies to prepare future engineers to lead in data-driven innovation and adapt civil infrastructure systems to evolving societal challenges. This course was developed by Neda Masoud, an associate professor in the department.
Civil and Environmental Engineering is also exploring a foundational AI course that would familiarize students with Python, AI fundamentals, ethical AI use and applications of generative AI, laying the groundwork so that the courses above could skip the introductory material.
Biomedical Engineering offers the senior-level course “AI in BME” (BIOMEDE 487). It focuses on practical applications of AI in biomedical engineering with hands-on tutorials. Students learn to use machine learning tools; integrate them with biomedical datasets including imaging, omics and health records; and apply these approaches to treating infections, cardiovascular and neurological diseases as well as cancer. It was developed by Sriram Chandrasekaran, an associate professor in the department.
Computer Science and Engineering recently launched a graduate level course, “AI for Science” (CSE 598), which is designed to welcome students from all disciplines who have previous experience with machine learning. It was developed by Alexander Rodríguez, an assistant professor in the division. The course focuses on machine learning and deep learning techniques. It includes mathematical modeling, neural networks and foundation models as well as topics that help students address AI’s weaknesses, such as determining uncertainties in its answers and the inferences that led to those answers.
Electrical and Computer Engineering is launching three new graduate courses under number ECE 598, that home in on specific applications.
Naval Architecture and Marine Engineering offers a sub-plan (concentration) in autonomous systems for masters of science in engineering students. Courses include:
Mechanical Engineering’s “Computational and Data-Driven Methods in Engineering” (ME 599-001), led by Xun Huan, an associate professor in the department, focuses on the underlying concepts that power AI and machine learning, including probability, regression and inference. The aim is to help students understand how machine learning and foundational models work, equipping students to be thoughtful and responsible AI users in their engineering careers.
Chemical Engineering has been developing, “Applied Data Science for Engineers” (CHE 696), led by Bryan Goldsmith, an associate professor in the department. This course prepares students to use data science and machine learning tools with a central focus on tools used in engineering and science applications. This includes Python programming, data curation and supervised and unsupervised machine learning, such as neural networks—culminating in a data science project on a topic chosen by the student.
In Aerospace Engineering, “Statistical Inference, Estimation and Learning” (AEROSP 567) includes a grounding in probability theory, sampling algorithms, estimation and statistical inference. Guided projects translate theory into practice: estimating the probability of rare events that may cause system failures, designing experiments to collect as much information as possible to train a model of a system, solving inverse problems to identify the source of a satellite’s attitude drift, and building state estimators that can be used to integrate sparse hypersonic-vehicle telemetry in real time.
By deriving and coding every algorithm themselves, participants gain a true command of the fundamentals that power modern machine-learning, positioning them to develop reliable, interpretable solutions rather than relying on black-box tools, according to course developer Alex Gorodetsky, an associate professor in the department.
“Complex Systems Design and Integration” (AEROSP 740-003) expects students to arrive with coding experience and includes a substantial section on statistical methods and data-driven modeling. This part of the course introduces techniques often used in machine learning—such as regression, surrogate modeling and neural networks—as methods for design exploration and optimization of complex engineered systems. The course was developed and is taught by Gökçin Çınar, an assistant professor of aerospace engineering.
In Nuclear Engineering and Radiological Sciences, Radaideh also teaches “Applied Machine Learning for Nuclear Engineers” (NERS 590) in which students work with real-world nuclear applications in fission, fusion, reactor control, security and medical imaging. The course combines weekly machine learning theory lectures and programming labs, covering topics such as deep learning, neural networks, reinforcement learning, generative AI and large language models. Students use these foundations to apply machine learning to a problem of their choice.
People who graduated years or decades before ChatGPT splashed onto the scene may also be looking for ways to harness AI as engineers, as well as students in other areas of study, and Michigan Engineering professors aim to serve these needs with shorter or more narrowly focused learning opportunities.
Biomedical engineering professors are developing a series of online modules to teach foundations of programming machine learning and AI, such as linear algebra, in applied contexts. Chandrasekaran; Zhongming Liu, an associate professor in the department; and Anne Draelos, an assistant professor, are leading this effort.
In particular, they aim to provide flexibility and adaptability across different educational backgrounds. The modules will cover fundamental AI and machine learning concepts as applied to specific biomedical engineering challenges, such as applications in neuroscience, imaging and systems biology. These modules are intended for graduate students and professionals, to be updated as AI continues to evolve, with an eye to creating a certificate program.
A certificate program offered through the Center for Connected and Automated Vehicles at U-M, provides a 5-module course intended to prepare undergraduate and graduate students to help develop AI for self-driving vehicles. It includes modeling and machine learning methods in addition to topics like cybersecurity, dynamics and control, and legal considerations.
Teaching beyond U-M, mechanical engineering professor Wei Lu offers a series of three massive open online courses (MOOCs) on Coursera, collectively called AI for mechanical engineers. The modules are AI for autonomous vehicles and robotics, AI for design and optimization, and AI for energy and biomedical applications. Coursera reports more than 1500 enrollments for the series as of publication.
U-M institutes also engage academic professionals with opportunities to learn about AI. The Michigan Institute for Computational Discovery and Engineering hosts an annual workshop on scientific foundation models, and the Michigan Institute for Data Science offers a series of mini-symposia on AI, the last featuring many engineers who are using AI in research.