At the Electrical and Computer Engineering Department Heads Association (ECEDHA) 2026 conference, I had the chance to sit down with several professors to talk about a question that’s on a lot of our minds right now: how do we teach AI (especially edge AI) in electrical and computer engineering programs without losing the fundamentals that make engineers effective? Across these conversations, a few common themes emerged. First, AI is no longer optional for ECE students. However, integrating it into the curriculum is messy, context-dependent, and still evolving. Faculty are wrestling with when to introduce AI tools, how to balance theory with practice, and how to prepare students for an industry that increasingly expects both software and hardware fluency.
Sid Deliwala: Teaching Edge AI Through Project-Based Learning
Sid Deliwala from the University of Pennsylvania takes a top-down approach by introducing AI concepts early (freshman or sophomore years) through project-based learning. Rather than waiting for students to build a full theoretical foundation, his approach is to get them building quickly using tools like Raspberry Pi, basic computer vision, and simple AI integrations. The goal is to spark curiosity and show what’s possible.
What’s particularly interesting is how this top-down exposure naturally leads students to discover the limits of their knowledge. When their projects hit performance bottlenecks (e.g. process video in real time) they begin to understand why hardware constraints, optimization, and deeper theory matter. This creates a strong motivation to engage with more advanced material later in the curriculum. He also emphasizes that hands-on labs, group projects, and practical assessments may become even more important in a world where AI can easily generate code or written answers.
Omiya Hassan: Bridging AI and Hardware in Engineering Education
Omiya Hassan from Boise State brings a more hands-on, curriculum design perspective, especially around edge AI systems and hardware acceleration. One of the biggest gaps she sees in students entering graduate programs or industry is a lack of understanding of how AI actually works (at the model level and at the hardware level). ECE students often lack exposure to machine learning, while computer science students may not understand the hardware their models run on.
Her approach is to explicitly bridge that gap by pairing students from different backgrounds and structuring courses to move from software concepts (models, training, evaluation) into hardware concepts (memory, compute, accelerators). This cross-disciplinary collaboration mirrors what happens in industry and helps students learn not just technical skills, but how to communicate across domains. Another key takeaway is her nuanced stance on LLMs: they can be powerful tools in advanced courses, but their use should be intentional and aligned with learning goals, not a shortcut around understanding.
Mario Simoni: AI in Electrical and Computer Engineering
In this conversation with Mario Simoni, Department Head at Rose-Hulman Institute of Technology (RHIT), we focused on the big-picture challenge: how AI fits into the broader ECE curriculum. One of the key takeaways is that ECE’s role in AI is not just about using models, but rather it’s about understanding how those models interact with physical systems. That includes everything from running inference on resource-constrained hardware to designing energy-efficient infrastructure that can support large-scale AI workloads.
A major tension he highlights is the balance between teaching foundational skills and introducing powerful tools like LLMs. If students rely too heavily on AI too early, they risk never developing the intuition needed to be effective engineers. But if programs ignore AI entirely, graduates will be unprepared for industry. His suggestion (and one I heard throughout the conference) is a staged approach: restrict AI use in early courses where students are building fundamentals, and gradually introduce it as a tool once they have something meaningful to “multiply.”
Conclusion: Finding the Balance
Taken together, these conversations highlight that there’s no single “right” way to integrate AI into ECE education. Instead, we’re seeing a spectrum of approaches (from bottom-up fundamentals-first models to top-down, project-driven exploration) with most educators trying to find a balance somewhere in between. What seems clear is that AI should be treated as both a subject of study and a tool, introduced thoughtfully based on where students are in their learning journey.
For educators, a few themes stand out. First, fundamentals still matter. Second, edge AI provides a natural home for ECE programs by connecting algorithms to real-world constraints like power, latency, and memory. Finally, hands-on, interdisciplinary learning is becoming increasingly important, both to mirror industry practice and to give students a deeper, more intuitive grasp of the systems they’re building. We’re still early in figuring this out, but conversations like these are a good step toward shaping what the next generation of engineering education might look like.