Open-source software is foundational to modern ML — from inference engines and distributed computing runtimes to agent frameworks, training systems, and evaluation tools. Yet the research community and the OSS community often operate in parallel: researchers depend on OSS tools daily but rarely engage with their development, while maintainers struggle to connect with the academic users who could most benefit from — and contribute to — their projects.
This social aims to bridge that gap. We bring together OSS maintainers, contributors, and researchers across institutions and projects for an evening of informal exchange.
Meet the maintainers behind the tools you use daily. Discover pathways to contribute — even as a busy academic — and find collaborators who share your infrastructure challenges.
Engage directly with your academic users. Share the challenges of sustaining open infrastructure, recruit contributors, and find researchers who can validate and co-develop your roadmap.
Lower the barrier to OSS contribution. Connect with project leads, ask candid questions, and find where your skills can make an immediate impact in real-world ML systems.
Strengthen the reproducibility and openness of ML research by deepening ties between the people who build the tools and the people who use them in science.
Maintainers and contributors from across the ML open-source ecosystem will be present. Projects span the full stack of modern ML infrastructure.
Maintaining an ML OSS project and want to participate? See the Call for Talks below — all project maintainers are welcome.
The projects above are just a starting point. We actively want to hear from any open-source community working on ML infrastructure — inference, training, distributed systems, agents, evaluation, data pipelines, RL frameworks, multimodal stacks, or anything else that shapes how modern ML gets built. If your community is not on the list yet, this section is for you.
Lightning talks are short (3–5 minutes), informal, and intentionally low-overhead. No slides required, no polished pitch expected — just bring what you're building, what's hard about it, and how the research community can help. We especially welcome early-stage projects, academic OSS efforts, and maintainers from underrepresented regions and institutions.
Any maintainer, contributor, or steward of an open-source ML-related project. Independent developers, academic labs, industry teams, and community-run projects are all equally welcome.
Email the organizers a one-paragraph description of your project and what you'd like to share. Submissions are reviewed on a rolling basis until the lightning-talk slate is full.
A lightly structured evening designed to maximize genuine connection — not slides decks.
Brief framing of the social's goals and why we're here.
Fast-paced 3–5 minute talks from a diverse set of open-source project maintainers — covering what they're building, what challenges they face, and how the research community can help.
Shared themes, cross-project questions, and community-wide challenges. Lightly moderated, attendee-driven.
Open mingling with project tables. Find the maintainers you've been meaning to talk to. Light refreshments provided.
Works at the intersection of operations research and AI systems, from software to hardware, with a focus on accelerating and reducing the cost of LLM inference through online optimization, scheduling, queueing, and resource allocation.
jerryzhou@ust.hkCore contributor to VeRL-omni and vLLM-Omni — multimodal RL training and inference systems for large foundation models.
huang.yongxiang2@huawei.comBridges university research, open-source communities, and product engineering across AI compilers, model training and inference, and data+AI infra
jin.fengmei@huawei.comOpen to all ICML 2026 attendees — researchers, students, maintainers, and anyone curious about ML open-source infrastructure.
Registration is open — sign up via the form above. Contact the organizers if you'd like to give a lightning talk.
Anyone at ICML 2026 with an interest in ML open-source software — whether you're a researcher who uses OSS tools, an active maintainer or contributor, a student exploring how to get involved, or simply curious about the infrastructure underpinning modern ML research.
Not at all. We actively want researchers and students who are users of OSS tools, as well as people who are curious about contributing for the first time. The goal is to mix both communities.
Yes — we welcome maintainers and contributors from any ML-related open-source project. Talks are 3–5 minutes. Reach out to the organizers via email or the registration form to express interest.
No. The social is free to attend for all registered ICML 2026 participants. Light refreshments will be provided.
The social will take place at the main ICML 2026 venue — COEX Convention & Exhibition Center, Seoul, South Korea, in Room E1-E4.
The social runs 19:00 – 21:00 (2 hours) on Wednesday, July 8, 2026, during ICML 2026 week.