From Research Project to Open Source Ecosystem: Bring Your Academic PyTorch Project to PyTorchCon NA
Some of the most interesting work being built with PyTorch starts in universities, research labs, student groups, and academic institutions. A new model architecture. A library created to support a paper. A benchmarking framework. A research tool that solves a problem no existing project addresses.…
Some of the most interesting work being built with PyTorch starts in universities, research labs, student groups, and academic institutions. A new model architecture. A library created to support a paper. A benchmarking framework. A research tool that solves a problem no existing project addresses. An educational toolkit developed for a university course. A system that begins with three graduate students and unexpectedly finds users around the world. What happens after the research project works? The paper may be published. Students graduate. Grants end. Research priorities move on. And a project with genuine value to the broader PyTorch community can struggle to make the transition from research artifact to sustainable open source project. The PyTorch Foundation OSPO & Academic Outreach Working Group wants to help bridge that gap. As part of Day 0 programming ahead of PyTorch Conference North America 2026, we are hosting an Academic Workshop focused on Education & Community Adoption. And we are inviting academic projects worth highlighting across a range of domains to submit using the “PyTorch Projects: From Academia to Adoption” GH template in the OSPO and Academic Outreach Repository. Call for Academic PyTorch Projects We invite students, researchers, faculty, research engineers, university labs, Academic OSPOs, and research institutions to submit open source projects created within academia, substantially developed by academic contributors, or emerging from academic research using PyTorch, DeepSpeed, Helion, Ray, Safetensors, and vLLM. Eight (8) selected projects will be invited to present short lightning talks, five (5) mins each, during the workshop. Presenters must be able to attend in-person to present their projects. Following the talks, project teams and other meetup attendees will join a hands-on workshop and mentor office hours with maintainers, open source program managers and professionals working in industry and research organizations, open source contributors, and members of the broader PyTorch community. The goal is simple: Help promising academic projects take their next step toward becoming sustainable community projects. What Kind of Projects Are We Looking For? We are particularly interested in projects where the work has demonstrated technical value beyond a single experiment or paper and could be useful to a wider community. Your project needs to use PyTorch or one of the other PyTorch Foundation hosted projects and might be: a library or developer tool; a research framework that extends one of these projects for a particular domain; a compiler, runtime, profiling, distributed training, inference, or optimization project; work in computer vision, NLP, robotics, scientific computing, healthcare, responsible AI, accessibility, or another research area; an evaluation or benchmarking framework; a reproducibility toolkit or research infrastructure project; an educational tool that makes these projects easier to teach or learn; an open source implementation accompanying significant academic research; a project connecting these projects with other parts of the open source AI stack; or something we have not thought of yet. You Don’t Need to Have Everything Figured Out Research software vs. sustainable, community-driven open source projects are often developed under very different conditions. Academic teams often optimize for experimentation, publication, reproducibility, and research velocity. Sustainable open source projects additionally need to think about documentation, releases, governance, security, contributor onboarding, testing, licensing, maintainer succession, and community operations. Those gaps should not prevent good academic work from reaching a wider community. The OSPO & Academic Outreach Working Group was created in part to help academic projects navigate exactly this transition. So perhaps your project has excellent research behind it, but: the documentation needs work; new contributors find the repository difficult to navigate; only one person understands the release process; the original student maintainer is graduating; governance exists informally but is not documented; CI or packaging needs improvement; there is no CONTRIBUTING.md, SECURITY.md , or maintainer policy; you are unsure how to build a contributor community; users are appearing, but you do not know how to turn users into contributors; or you simply do not know what would be required to eventually become a PyTorch Ecosystem project. The workshop is designed to help identify those gaps and turn them into an actionable roadmap. Workshop Program Overview 1. Academic Project Showcases Selected projects will present short lightning talks covering: The problem they are solving, How they use or extend PyTorch Foundation hosted projects, The research or academic context behind the work, Who is using the project today, Why the project could matter to the broader community, What support, resources, or changes could help the project reach its next stage 2. Maintainer & Industry Open Source AI Leader Office Hours After the lightning talks, we move from show-and-tell to show-and-build. Project teams will work with maintainers, PyTorch ecosystem contributors, professionals who lead and support open source ecosystems within companies and other organizations, and community members around actual repositories. Depending on the project, conversations may cover: Open source readiness: Is it obvious what the project does? Can someone install it, run it, test it, and understand how to contribute? Governance: Who makes decisions? Who can merge changes? What happens when the original academic maintainers move on? Reproducibility: Can another researcher reproduce the results? Are datasets, benchmarks, evaluation methodology, dependencies, and artifacts clearly documented? Community: How does a new user become a contributor? Where are decisions discussed? Are issues approachable by people outside the original lab? Security and maintenance: How are dependencies, vulnerabilities, releases, and supported PyTorch versions handled? Ecosystem positioning: What is the project’s relationship with PyTorch? What unique problem does it solve? What would need to change before a future PyTorch Ecosystem application made sense? Our ambition is that teams leave the workshop with a prioritized set of next steps for making their project easier to use, contribute to, maintain, and grow. Who Should Submit? We encourage submissions from: university students and graduate researchers; professors and academic educators; research labs and institutes; research software engineers; Academic OSPOs; maintainers of research-originated open source projects; and academic communities building tools around the PyTorch Foundation hosted projects. Projects can come from anywhere in the world. You do not need to represent a large institution, have thousands of users, or already have a mature open source organization behind you. We care about the idea, its usefulness, its relationship with PyTorch and the other PyTorch Foundation hosted projects, the work already happening around it, and its potential to grow beyond its original academic environment. Evaluation Criteria Projects will be evaluated based on: Academic relevance: The project originated in, is substantially developed by, or supports academic research Ecosystem relevance: The project uses, extends, or integrates with PyTorch or another PyTorch Foundation-hosted project Open source readiness: The project has a public repository, an open source license, and enough documentation for mentors to review it Potential impact: The project addresses a meaningful problem and could benefit users beyond its original research team. Team commitment: At least one maintainer can participate in the event and continue working on the agreed next steps afterward. Cohort diversity: The final selection should represent a range of institutions, research areas, project stages, and geographic regions What to Include in Your Submission Tell us: What is the project? Give us the repository and a short description. Where did it come from? Tell us about the university, lab, research group, course, or academic collaboration behind it. How does it use or extend PyTorch or its ecosystem? Who is it useful to? Researchers? Students? ML engineers? A particular scientific community? What exists today? Users, contributors, releases, papers, deployments, courses, benchmarks, or other evidence of adoption are all useful context. Where are you stuck? Documentation? Governance? Packaging? Testing? Contributors? Maintainers? Visibility? Reproducibility? Ecosystem readiness? What would you like this workshop to help you achieve? We are not looking for perfectly polished applications. We are looking for projects worth helping. Bring Us the Projects Academia Shouldn’t Lose There is extraordinary open source AI work being created inside universities and research institutions. If you have built something with PyTorch that deserves a life beyond the paper, thesis, course, grant, or research group that created it, submit it to the PTCNA Day 0 Academic Workshop. Submit your academic PyTorch project → PyTorch OSPO and Academic WG Issues (New Issues→ PyTorch Project : Academia to Adoption Submission deadline: September 30th 2026, midnight PT Registration for this event is separate from the PyTorch Conference and will be open later this week. Learn more about the PyTorch Foundation OSPO & Academic Outreach Working Group: https://github.com/pytorch-fdn/wg-ospo-and-academic-outreachSource: PyTorch — Published — Category: Open Source