AI-Enhanced Project-Based Learning in Large Cohorts Courses: Embedded Systems and Beyond

Implementation of the Project/Course

The Embedded Systems course at ETH Zurich is a core mandatory course with more than 380 students from computer science, electrical engineering, and mechanical engineering. Unlike small, dedicated PBL courses, this course scales project-based learning to a very large, heterogeneous cohort. Because embedded systems by nature require hands-on experience, the course integrates lectures with extensive project-based sessions. To achieve this at scale, a custom educational hardware platform was created, offering sensors, processing, and actuation, so that every student team can design, test, and demonstrate real prototypes. Roughly 50% of the course is devoted to lectures, while 50% focuses on active, hands-on labs & projects. Students apply concepts directly on the hardware platform, solving authentic engineering challenges. Both the teaching team and an AI assistant support students, but autonomy and discovery are central.

Students receive continuous, multi-channel feedback:

• AI Assistant: Fine-tuned on course is trained with hardware documentation, past exams, student feedback, and especially PBL workflows, it delivers real-time guidance for debugging, design, and even communication (video pitch preparation).
• Human Instructors and Tutors: Provide weekly feedback through labs and milestone reviews.
• Peers: Structured peer evaluations promote reflection and mutual learning.

Engagement is built around real hardware prototyping. The platform supports both entry-level and advanced projects, ensuring inclusivity in a diverse cohort. Single Student or eams of 2–3 students work on open-ended challenges and present both functioning systems and short pitches, creating high motivation and ownership.

Lectures and labs allow student-teacher interaction; Piazza and Moodle support student-student exchanges; teaching staff coordinate through weekly syncs. The AI assistant ensures 24/7 support, bridging gaps outside teaching hours.

Lectures and milestone reviews are synchronous; project development and AI-guided work are asynchronous, allowing students to learn and iterate flexibly.

Support is multi-layered: (1) hardware platform access, (2) lab mentoring, (3) peer collaboration, and (4) AI guidance. This ensures scalable yet personalized support.

The assessment combines traditional and innovative elements: Prototype demonstration using the hardware platform. Short video pitch (<2 minutes) where students explain their project and results, demonstrating communication skills and applied understanding. Targeted exercise on energy and latency analysis, directly tied to the student’s project, reflecting critical real-world metrics in embedded systems. Final exam aligned with project content.This mix ensures students are evaluated on both technical and transversal skills.

Scaling hands-on PBL to a large core ETH course was the main challenge. The custom hardware platform ensured every student could experience true hardware/sof very student could experience true hard.

Motivation, Project Mission, Vision Statement

Embedded systems power technologies from medical wearables to autonomous robots. Teaching them effectively requires both theory and hands-on practice, which is challenging in large, diverse cohorts. Our mission is to ensure every ETH student, regardless of background, gains real experience with hardware/software integration in a core course.

We built an open, modular hardware platform enabling scalable prototyping, and complemented it with the first AI assistant tailored to PBL education. Students learn by building prototypes, presenting them via video pitches, and analyzing key real-world metrics such as energy and latency.

The project values inclusivity (supporting both novices and advanced learners), scalability (making PBL feasible for 400+ students), and sustainability (tools transferable to other contexts). Importantly, the course functions as a testbed for innovative teaching methods, proving that hardware-supported, AI-enhanced PBL can be applied in large mandatory courses and smaller specialized ones. This model has already been piloted with success at Ashesi University in Ghana, showing its adaptability to less resourced institutions and reinforcing our vision: to set a new standard for engineering education where AI and hardware co-design empower students globally to become creative, independent, and industry-ready innovators.

Innovative Elements

This project pioneers three innovations per se that are even combined together :

1. An open, modular hardware platform integrating sensing, processing, and actuation, enabling scalable hands-on projects across fields such as biomedical systems, robotics, and wearables.

2. A PBL-based redesign of a large and important ETH course, with assessments including prototypes, video pitches, and energy/latency analysis.

3. A custom AI assistant, fine-tuned with course materials, hardware documentation, and PBL workflows, providing real-time support for debugging, design, and communication.

Implemented at scale, the course serves as a testbed for transferable approaches across ETH and beyond. Its adoption at Ashesi University in Ghana demonstrates that the model is effective not only in a leading technical university but also in less resourced contexts, highlighting its global potential.

Effects on Student Learning

To enable our innovations, we significantly restructured the course: redesigning labs and exercises, adapting hardware, revising exams, and embedding the final PBL project. We monitored student outcomes during the transition (2023–2024). In 2024, when our innovations were fully integrated, evaluations showed clear improvement: an average of 4.6/5 and 4.9/5 on the labs ETH evaluatons. Overall course appreciation rose by nearly 1 full point compared to previous year. Evidence of effectiveness is also seen in student demand: the 2025 edition reached a record 482 registrations before semester start, with 82 students on the waiting list, many referred by peers and colleagues. Students show stronger engagement, critical thinking, and knowledge transfer, supported by the AI assistant and authentic hardware prototyping, and we have feedback of this from student association and student prjects such as ARIS, ForzaETH, Swissloop that claims thanks to the ES coruse and the better level of it.

Which Elements of Your Project Would You Recommend to Others?

Several elements of this project are highly transferable and scalable. The open hardware platform can be easily adopted in other courses, from small project seminars to large core lectures, enabling authentic hands-on learning with sensing, processing, and actuation. The AI assistant is modular and adaptable, offering real-time support that can be retrained for different domains, and focus on PBL. The PBL structure with prototypes, short video pitches, and performance metrics (e.g., energy and latency) has proven effective for engagement and assessment. Together, these elements provide a tested model for scaling PBL to large cohorts, while also working in smaller or less resourced contexts, as demonstrated by its successful use at Ashesi University in Ghana and workng other courses.