Beyond Workshops: RSE-Led Hackathons as Experiential Learning for Software Practice in the AI Era

Abstract

Research software development is becoming increasingly decentralized. Domain scientists, data analysts, and research staff who are self- or semi-taught programmers are writing more code than ever, and the rise of AI coding assistants has dramatically lowered the barrier to producing software artifacts. But the fundamentals of good software engineering (like version control, testing, modular design, documentation, collaborative workflows) have not changed. AI tools, when used without guidance, will generate mountains of plausible-looking but poorly structured, untested, and unmaintainable code. The challenge for the research computing community is no longer just producing software but helping their community produce software that is reliable, reproducible, and sustainable. Research Software Engineers are uniquely positioned to address this challenge. RSEs inhabit the intersection of research practice and software craftsmanship. They understand both the domain pressures that drive researchers to "just get it working" and the engineering disciplines that make software trustworthy over time. But traditional approaches to spreading good practice like workshops, tutorials, learning materials or documentation can only go so far. They convey knowledge, but they rarely produce the kind of durable learning that comes from hands-on experience under real constraints. We argue that the hackathon format, when deliberately designed around inclusion and learning rather than competition, offers RSEs a powerful vehicle for experiential software engineering education. We present a case study from Dartmouth College, where the RSE team organized an internal hackathon that brought together staff from across Research Computing & Data, including systems administrators, research facilitators, data scientists, and software engineers, to collaboratively build Stepify, a functional AI-powered video-to-tutorial converter, in a single day. Three design principles guided the event. First, a job for every role: the project scope included UI/UX sketching, test case curation, output evaluation, and documentation alongside code, so that non-coders contributed meaningfully. Second, a sprint, not a race: we removed prizes, judging, and pressure to ship, framing the day as focused collaboration. Third, a prepared mind: RSE arrived with a preliminary architecture but presented it as a starting point, not a specification, leaving room for the team to reshape the design. The results were more impactful than just the working prototype. Staff who do not identify as software engineers reported feeling more confident engaging with development workflows. The event seeded ongoing conversations about code review, collaborative development, and software quality within the organization. As one participant reflected: "People consistently surprise themselves when they are trusted to try." This talk will share the hackathon design, its outcomes, and practical guidance for RSE teams looking to catalyze software engineering culture within their organizations by creating structured opportunities to apply best practices instead of lecturing about them.