Generating Future AI Leaders

Generative AI can produce a variety of interesting outcomes—some good, some dangerous.  

When Joshua DeAndria teaches students in the generative AI course in Northwestern Engineering's Master of Science in Artificial Intelligence (MSAI) program, his focus is on the more serious side. 

“No one will be harmed if your generated cat image has five legs,” he said. “The stakes are much higher in healthcare, finance, politics, and more.”  

This truth—and the weight of it—guides Joshua's approach to teaching about a technology that is becoming more ingrained in businesses every day. Deloitte's State of AI in the Enterprise report, published in January, reported that 80-90 percent of new use cases of AI in businesses involved generative AI.  

"We need to train our next generation of generative AI engineers," Joshua said. "People don’t just show up in industry full-fledged and ready to go. It takes time, intention, and resources to become a professional in this space." 

That's what the MSAI program strives to do—and it's Joshua's primary focus with his generative AI course. He knows there are plenty of resources online that students, and the public at large, could review to learn about generative AI, but Joshua believes there is far more value in learning it in his classroom. 

"When you’re learning in a silo you rarely challenge yourself," he said. "You skim formulas. You don’t do all of the practice problems. In short, there is less of a sense of accountability. My students are held accountable to learn—in as fun and rewarding a way as possible—and as a result they are more likely to actually learn."  

His knowledge, coupled with the expertise of his fellow MSAI faculty, is an added differentiator. 

"We all have experiences in industry and research that we can relate to students," he said, "and aim to tailor our materials based on student interest."  

Joshua designed his course to provide a comprehensive overview of generative AI techniques, with a focus on image and text generation.  

The class covers scalable infrastructure, variational autoencoders, generative adversarial networks, diffusion models, autoregressive models, multi-modal models, and world models.  

The course also includes hands-on projects and demonstrations to reinforce the lessons Joshua teaches. 

“Being at Northwestern has its perks,” Joshua said. “That includes working on Quest, our high-performance computing cluster that the entire class has access to throughout the quarter for their projects and our classroom demos.” 

In addition to Generative AI, Joshua also teaches the program's foundational data science course, where he emphasizes the importance of linear regression and tree-based models.  

"In my experience, the best models for many applications are the simplest and oldest ones," he said. "These models remain much more interpretable, easier to train, faster and cheaper than state of the art model architectures." 

Joshua also teaches in Northwestern's MBAi Program, a joint-degree program offered between the Kellogg School of Management and Northwestern Engineering.  

Prior to joining Northwestern in 2024, Joshua worked in the health tech space. He incorporates lessons learned from his professional experience into the classroom, but he also emphasizes the importance of students learning from one another. 

"I try to ask more questions instead of delivering a three-hour monologue," he said. 

Joshua continues to be impressed by the students he interacts with. His MSAI students fascinate and inspire him with their curiosity and creativity. 

"MSAI students are very interested in how the technical details can be expanded on or added to their toolkit," he said. "They want to know how everything works and how they can improve a process or try something that hasn’t been done before."