Machine Learning skills for all students with learning-by-doing bite-size workshops

The project focuses on practice-oriented workshops in machine learning (ML), aimed at all BSc, MSc, and PhD students at ETH. The goal is to provide students with a fundamental understanding of modern digital tools. This particularly includes Large Language Models (LLMs). The focus is on practical application and the safe use of these technologies. The project also aims to raise awareness of current challenges, such as biases embedded in models or the dependency on proprietary software. All teaching materials created will be made freely available as open source so that other departments and lecturers can also use them.

At the Student Project House, a sharp increase in demand for content and support in the field of AI has been observed. For example, in the Spring Semester of 2023, nearly half of the newly registered projects were based on AI technologies, although students often had little prior knowledge. Furthermore, analysis of the «Digital Makerspace open hours» showed that over 30% of inquiries related to machine learning and LLMs, making it the most common topic. Due to this demand, an initial prototype workshop on LLMs was developed and successfully conducted; all 30 spots were filled immediately without any advertising, which demonstrated the students› significant interest. Feedback from the participants showed a desire for follow-up workshops.

We designed five workshops for an average of 10-30 participants. Each 2 hour workshop began with a short 15-minute introduction into the topic, followed by practical exercises for the rest of the course. At the end of each workshop, we gathered feedback and checked the demographics of our students. One of our goals for each workshop is also, that the students can go home with something that they made and experience a success in learning.

For the physical setup, we found round tables, accommodating 4-6 students per table group, allows for more student-student interaction compared to a lecture-style approach.

In 2023 and 2024, we created three workshops: «Chat with your PDF», «AI Fairness» and «Computer Vision». The later two seemed to be less interesting for students. To address this, integral parts of these (less popular topics) were integrated into the «Chat with PDF» workshops in a shorter, more digestible format and the original workshops were discontinued. By 2025, «Chat with PDF» became already obsolete due to the rise of readily available tools. We then changed the workshop format again, based on student feedback, the workshop formats evolved to «Make your first web app with AI», «Make your own chat bot and self-host it». We suspect, that in the future we must remain flexible in choosing interesting topics.

Project goals for student engagement and diversity were largely achieved through continuous adaptation. For instance, the once-popular «Chat with PDF» workshops became obsolete by 2025 as many free tools with this functionality emerged. This prompted a shift to more practical, in-demand formats like «Make your first web app with AI» and «Make your own chat bot and self-host it». This practical learning was enabled by the creation of two key services, provided by us: a free LLM API key for modern AI models (funded by SPH donations), and a student web hosting platform that was established after ETH Zurich discontinued its own. This crucial, self-hosted, infrastructure directly supported the new workshop content.

Analyzing student feedback revealed that topics like «Make your first web app with AI» were more appealing to «Non-Computer-Science-Students». This shift had a measurable impact on diversity, aligning with our goal of empowering students from all fields. The female participation surged from 29% to 51% in the corresponding period. This demonstrates that aligning offerings with student interest and providing accessible tools effectively broadens reach and inclusivity.

These lessons highlight that in today’s rapidly changing environment, a flexible curriculum responsive to feedback and technological shifts is mandatory. Providing foundational resources, such as free API access and self-hosting capabilities, is a key strategy to empower students far beyond traditional technical fields. We will continue with offering our workshops to students. Having the innovedum grant helped us to pay staff who helped to prepare for the workshop and assist in the workshop. It was also helpful in setting up our infrastructure that was crucial for our woskhops.

The following goals were accomplished:
– Created a portfolio of 5 hands-on workshops on machine learning and specific use-cases of large language models and computer vision: «Chat with your PDFs, AI Bias, Computer Vision, Make your own chat bot and self-host it, Make your first web app».
– Delivered each workshop at least once per semester with the popular workshops being delivered between 2 and 4 times per semester.
– Documented each workshop and make it readily available online and open source so it can be reused by any ETH department and lecturer autonomously: the content can be found on https://gitlab.ethz.ch/sph/digital-makerspace/innovedum1
– All lecturers of ETH were welcome to attend the workshops if desired via public eventbrite.
– A total of 18 workshops were delivered at the Student Project House.
– A total of 265 workshop participations ~80% unique users.