The variety of programs and courses reveal the field is wide open.
Training for artificial intelligence (AI) engineers differs widely, indicating that norms and requirements are still in the process of being set. Generally, AI engineers have the job of designing, deploying and maintaining AI models to ensure operations in different fields are efficient and safe. This requires a background in machine learning (ML), statistics and programming. An AI engineer must know what data to collect, as well as how to utilize ML libraries like scikit-learn.
One of the common features of AI engineering-related academic and professional certification programs is the opportunity to apply skills learned in courses to specific problems. For example, an AI program may give a student a chance to design a method to cope with memory limitations of IoT-enabled devices like drones. Requirements to complete AI engineering-related professional certificate programs offered by IBM, MIT, Duke University and John Hopkins University typically include the ability to program in Python and possessing a good understanding of linear algebra and probability.
One of the first academic programs in AI engineering, a Master of Science in AI Engineering (MSAIE) at Carnegie Mellon University (CMU), showcases how universities are deepening and diversifying coursework in this area of study. Simultaneously, researchers in AI engineering at CMU’s Software Engineering Institute (SEI) are demonstrating through projects how to apply knowledge in AI engineering.
Students train according to their interest
CMU currently offers the MSAIE program at its primary campus in Pittsburgh, Pennsylvania. CMU-Africa, the university’s African campus in Kigali, Rwanda, offers a comparable degree, the Master of Science in Engineering Artificial Intelligence (MSEAI). The Kigali program, which is more generalized, is designed for students who intend to work in African countries.
“African countries do not have the technology infrastructure of Western countries. The students in the Kigali program are learning how to create and implement AI algorithms in areas where there are limited resources such as limited power. One of the focuses of the program is getting technology more widely distributed,” says Shelley Anna, the associate dean for faculty and graduate affairs and strategic initiatives at CMU’s College of Engineering.
The Kigali program started in the academic year 2021-2022 and has seen an enrollment of 21 students each year. The Pittsburgh campus has seen an enrollment of 33 students in the academic year 2022-2023. Its students are spread out across seven disciplines of engineering: biomedical, chemical, civil and environmental, electrical and computer science, information security, materials science and mechanical. Nine of the Pittsburgh students are in mechanical engineering. Seven each are in chemical and civil and environmental engineering. Other disciplines of engineering have between three to five students. The Pittsburgh program lasts between three and four semesters, depending on the discipline.
Most of the students in the Pittsburgh program have a B.S. in an engineering field. A student is not required to continue on in the same discipline of engineering in which they earned their undergraduate degree. In the first three semesters, all Pittsburgh students take required core courses, including Introduction to Machine Learning for Engineers, Systems and Toolchains for AI Engineers, Introduction to Deep Learning for Engineers and Trustworthy and Ethical AI Engineering.
The Pittsburgh students are encouraged to get summer internships. They also get exposure to the corporate world when professors partner with companies regarding class projects. Typically, companies will suggest or co-develop projects for students with the professor. Employers are already expressing interest in the Pittsburgh program’s first class of graduates.
“This is because graduates from the Pittsburgh program are determining how AI algorithms can improve operations in engineering systems like chemical plants. Their classes are showing them what possibilities and constraints exist for their discipline,” says Anna.
A number of the Pittsburgh students will have the opportunity to work on class projects. A project may involve applying AI algorithms to the student’s engineering discipline. There are currently opportunities to do research on additive manufacturing, development and securing of wireless edge networks, and refinement of autonomous physical systems like autonomous vehicles.
In the future, CMU hopes to connect the AI engineering graduate students in Kigali and Pittsburgh. Recently, the two groups were in contact in mid-April, when the Pittsburgh campus hosted approximately 25 students from the Kigali campus.
Current professionals perform interdisciplinary work
At CMU’s Software Engineering Institute, researchers and engineers in the AI division explore methods and practices to advance AI engineering. Their goals are to help establish AI engineering as a discipline and meet the needs of the U.S. Department of Defense (DoD). The DoD has been the Institute’s primary source of funding since 1984. The SEI is one of 42 federally funded research and development centers (FFRDCs) in the U.S.
An FFRDC is a nonprofit, public-private partnership that performs research for the U.S. government. Ten FFRDCs are sponsored by the DoD. This explains why the Institute’s research centers on projects such as heightening cybersecurity, improving systems engineering for DoD agencies, and applying AI algorithms to increase safety for U.S. troops.
“AI engineering’s applications for DoD include use cases such as predictive maintenance, threat detection and battlefield healthcare,” says Carrie Gardner, an AI researcher at the Software Engineering Institute and a team lead in the Institute’s AI division.
Researchers in AI engineering also assist the DoD in other areas such as exploring next-generation software architectures, AI-optimized hardware design and test and evaluation standards. In 2020, SEI researchers provided feedback on two technology development programs at the Defense Advanced Research Projects Agency (DARPA). SEI researchers helped improve tools and designs for microelectronics production by sharing their input on efforts in DARPA’s Domain-Specific System on Chip (DSSoC) program and Software Defined Hardware (SDH) program.
Researchers in the AI division at the SEI have graduate degrees in a range of disciplines, including computer science, information science and electrical and computer engineering. The SEI conducts applied research and system implementation prototyping to surface practices, methods and tools for rigorous AI engineering standards.
“The realm of tasks for AI engineering at the SEI is wide. Researchers may investigate a fundamental challenge of AI implementation, such as patterns for auditing and interpreting AI output. Engineers may design, develop and field prototype AI systems – testing the readiness of technology implementations. Together, researchers and engineers surface resources to advance the state of practice for AI engineering,” says Gardner.
Work on DoD-sponsored projects may be sensitive. Yet the SEI’s mission includes transitioning research to the public.
“SEI researchers try to share as much as possible when it is appropriate. We write articles for peer reviewed journals, present at academic and DoD-related conferences, and give talks to CMU students and the public on topics like next generation architectural concerns for AI systems,” says Gardner.
The majority of researchers at the SEI are not CMU faculty members, and SEI researchers do not typically teach MSAIE classes. In addition, the SEI has a limited number of student interns. The interns are selected from a number of college programs in addition to the MSAIE program.
Yet the SEI is making efforts to establish AI engineering as an engineering discipline, much as it did for software engineering, starting in the 1980s, says Richard Lynch, manager of public relations for the SEI.
“We’ve published white papers on our three pillars of AI engineering. These are that AI should be human-centered, scalable, robust and secure. We’re also interested in how to develop an AI-capable workforce,” says Lynch.
SEI researchers’ close communication with DoD agencies has led to a shared understanding that paths to gain knowledge, skills and abilities in AI engineering include on-the-job training. For example, soldiers can use AI-enabled systems to identify threats on the battlefield. In order to perform such work, they must first learn how data collection will affect the outcomes of the system’s object detectors.
One of the phenomena that is bringing together students and professionals in AI engineering is the recent public conversation about generative AI. Generative AI is defined as algorithms that create new content like images and video in response to prompts.
“News about what generative AI is sharing makes it possible for us to hear from people at different skill levels, in different disciplines. The conversation is attracting people to the field. It’s also getting future and current AI engineers to discuss how we can comply with existing ethics rules and address new concerns,” says Gardner.