For decades, academic libraries were defined by what they archived. Today, they are increasingly defined by what they enable.

During a webinar, Sayeed Choudhury, Director of the Open Source Programs Office (OSPO) at Carnegie Mellon University (CMU), recounted a strategic assertion made to his university leadership: « Open source is the connective tissue of open science. » While open science encompasses diverse components like data, hardware, and publications, Choudhury says that open-source software is the common element linking them all together. Catch the entire 53-minute session here and take a look at the slides. This was the final webinar in the series “From Open Science to Open Source: A New Step for Libraries?”, organized by Software Heritage, Doranum with L’Atelier de la donnée en Normandie, and Média Normandie.

A center of competency with global reach

Founded in 2022 with support from the Alfred P. Sloan Foundation, the CMU OSPO was established to serve as a hub for managing, curating, and sharing the open-source software produced within the university. Choudhury describes the CMU OSPO as a « center of competency. » At a research powerhouse like CMU, expertise in engineering, licensing, and community engagement is everywhere, but it is often siloed. The OSPO’s mission is to build a community within the university where these experts can support one another.

Beyond research support, the OSPO fosters student engagement through initial fellowships and industry-partnered hackathons. Led by Tom Hughes, these efforts turn student needs into hands-on opportunities despite current university budget constraints. The office’s influence extends far beyond Pittsburgh. Led by industry veterans, CMU’s summer course on open source pairs students in Pittsburgh and Qatar—with future plans to expand to the campus in Kigali, Rwanda—with mentors from organizations like Red Hat and the Eclipse Foundation to gain hands-on experience in real-world coding.

Strength in numbers: The CURIOSS network

Choudhury emphasizes that the best way to make progress isn’t one university at a time, but by working across institutions. To that end, CMU is a key player in the Community for University and Research Institution OSPOs (CURIOSS).

Choudhury, an early originator of the network, notes that Curioss has grown into a thriving community of nearly 50 members across the US and Europe. Led by a dedicated team, the network serves as a vital resource for sharing information and collaborating on projects. For Choudhury, Curioss is a vital feedback loop; when CMU identifies a new pathway for project sustainability, those findings are shared with the network to help other institutions facing similar challenges.

Beyond commercialization: A new pathway for research

One of the most critical shifts the OSPO facilitates is the evolution of the technology transfer office (TTO). Traditionally, TTOs focus almost exclusively on the commercialization of intellectual property. However, Choudhury notes that commercialization is only one possible pathway for software.

The OSPO provides an alternative « project to product » migration path, helping labs move from a core group of researchers to a reliable, community-supported ecosystem. This is particularly vital for projects like Spiral (low-level signal processing) or Alice (a programming environment), which may not have a traditional commercial buyer but provide considerable public and academic value.

From the lab to the world

To illustrate how the OSPO helps researchers move from a « lab-based project » to a « community-supported product, » Choudhury highlighted five diverse projects currently being supported at CMU:

• Penrose: A high-level application that allows users to create complex diagrams by typing mathematical equations. While it began with simple diagrams, it has evolved into a tool capable of generating intricate visualizations, such as protein structures.
  
• Open energy outlook (Tomoa): Developed by the Scott Institute for Energy Innovation, this project is a large and complex model. It is specifically designed to provide insights into how various policies affect energy consumption.
  
• Spiral: In contrast to high-level apps, Spiral is a « very low level » infrastructure tool without a traditional user interface. It is used for deep-level engineering, specifically for writing signal-processing algorithms in storage systems.
  
• Alice: This project has a 30-year legacy as an interactive, game-like environment designed to make learning programming more engaging and accessible. It has reached a massive scale, with estimates suggesting that hundreds of thousands of students and teachers have used it to learn coding.
  
• DARE: Emerging from the Open Forum for AI, DARE serves as a gateway platform. It allows users to navigate and test different language models, prompts, and data settings within a single environment.

Setting the standard for open source AI

The OSPO’s work is also increasingly intersectional, reaching into the worlds of open hardware and artificial intelligence. Choudhury currently leads the Open Forum for AI (OFAI), an effort to define « open-source AI » in a way that is human-centered and transparent. This work has gained significant traction; the G7 technology and digital ministers recently released a vision piece that aligns with the community-led definition supported by the CMU OSPO.

Voices from the community: A conversation on reproducibility

During the interactive portion of the webinar, participants raised critical questions about the intersection of industry methods and academic goals.

Question from the floor: Industry often prioritizes rapid deployment through containers. In academia, our goal is often bitwise reproducibility over the long term. How does the OSPO balance these potentially conflicting goals?

Choudhury: There are some major differences between the way industry works with open source and things about open source …Very broadly speaking, the biggest difference is just the scale at which we operate. As you pointed out, when you’re working with enterprise-level software, containerization is a really important aspect. It’s different for an individual lab working on a tool that, quite frankly, they may not even work on after they’ve done their research or finished a grant. We’re very sensitive to those differences and understand we don’t really ever show up and say, « You must work the way industry works. » We tailor all of those conversations and consultations based on the needs of the individual projects.

It’s like a set of circles. The inner core consists of universal foundations—engineering, security, and licensing practices essential for any public software release. As you move outward from that center, requirements naturally diverge, and our process reflects and respects those differences…

Question: Were our OSPO activities linked in some way to the training provided to PhD students regarding, for instance, software development or not?

Choudhury: Absolutely. Most faculty don’t want to spend class time teaching Git or Python. By providing these options through the OSPO and our « Open at CMU » partners, we ensure that both undergraduate and PhD students understand the consequences of their coding choices, especially as AI-generated code becomes more prevalent.

Question: What’s your advice for finding all the code produced in a university and how to ‘valorize’ it?

Choudhury: I really like that word, valorize….It sounds like a simple problem. It is not….We started by doing simple searches in GitHub…which is just completely chaotic… I keep calling Carnegie Mellon CMU… but there’s a university called Central Michigan University, which also calls itself CMU. …We recognized very quickly that this is difficult. We then worked with an external group called NumFocus with Jonathan Starr… and we reached a point where we thought there were at least 5,500 or so open source projects of Carnegie Mellon.

…I said [to the provost], ‘Even if 1% of this information is valid. That’s 55 projects. Do you want all of those 55 projects to go figure out on their own what to do about sustainability and community engagement and governance?’ And the answer was, of course not. Go ahead and do this more systematically. …The Curiouss Network fortunately has a PhD student who did this and has released it publicly as open source software…[now] anybody has access to it. »

The path forward: A collective effort

Ultimately, the OSPO is an organizational framework designed to turn software into a permanent asset of the scientific record. By moving beyond the « lab-based project » model, universities can ensure their digital contributions remain accessible and functional long after a specific grant concludes.

For Choudhury, the success of this movement hinges on radical collaboration across borders and sectors. « The best way to make progress is not one university at a time, » he concluded. « It’s working across universities and working across nonprofits. » In this new model, the library becomes more than a repository; it becomes a launchpad for the code that will power the next generation of scientific breakthroughs.