AI Needs Its Linux: Oumi Comes Out of Stealth with an Open Source Vision 

A new company called Oumi has just been launched, emerging from stealth today with $10 million in seed funding. The company’s founders describe Oumi as the world’s first unconditionally open source AI platform, positioning it as a collaborative and transparent alternative to today’s proprietary AI ecosystems, similar to how Linux revolutionized operating systems.

AIwire spoke with Oumi co-founders Manos Koukoumidis and Oussama Elachqar to find out what this new platform means for the AI world.

Making AI Research More Accessible

Koukoumidis, Oumi’s CEO, explains: “What we're announcing is the first unconditionally open AI platform that will make it possible for the tens of thousands of AI researchers, scientists, and developers around the world to not just to more easily leverage and harness frontier AI, but also, more importantly, to collaborate on the same platform, working together to advance frontier AI in a collective way that will lead to more transparent and responsible development.” 

Oumi is developed in collaboration with 13 leading AI universities in the U.S. and the U.K., bringing together some of the most respected names in technology and innovation, including University of Illinois Urbana-Champaign, Carnegie Mellon University, Princeton University, Caltech, UC Berkeley, GeorgiaTech, Stanford University, University of Washington, New York University, MIT, University of Waterloo, University of Cambridge, and University of Oxford.

The initiative emerged from discussions with top academics, including Ruslan Salakhutdinov and Amit Acharya from Carnegie Mellon University, who highlighted the growing challenges AI researchers face. These include difficulty conducting large-scale experiments and integrating tools and infrastructure, even when resources like GPUs are available.

Oumi co-founders Oussama Elachqar and Manos Koukoumidis. (Source: Oumi)  

Koukoumidis says, “Salakhutdinov was telling me on a Sunday night how hard it is for his students to do this type of research. He said, ‘You know, the last few years have become a lot harder than it was before. We cannot contribute to the frontier anymore. When it comes to larger scale experiments, even smaller scale, it's too hard for students to figure out all the different things that need to be put together to do this type of research.’”

The founders then spoke with other top AI researchers and PhD students who also expressed concern about academia's diminishing role in AI innovation.

“What about the future of AI, if it's becoming a black box that we just interface with an API? That's not what academia should do. It's not how AI and the sciences should be advanced,” Koukoumidis relays.

Why AI Needs Its Linux

Oumi aims to be the Linux of AI, providing an open source alternative to the dominant, closed AI ecosystems. Unlike proprietary AI systems that function as black boxes, Koukoumidis explains that Oumi is built around open code, open data, open model weights, and open collaboration, making it easier for researchers to build upon each other’s work rather than struggling with fragmented, ad-hoc setups.

The founders also believe that AI research today is hindered by unnecessary complexity as top researchers and PhD students waste valuable time cobbling together infrastructure rather than focusing on advancing the science. By offering a unified, transparent platform, Oumi seeks to remove these barriers, much like how Linux revolutionized operating systems by fostering widespread collaboration and standardization.

Big Tech Won’t Dominate Forever

As we saw this week with the explosive news of DeepSeek and its impact on the AI landscape, the companies building the largest and most expensive AI models won’t necessarily be on top forever. 

“I think they're going to be trying to play catch up after a while with what's going to be happening in open source,” Koukoumidis says. “But to make this happen, open source needs its Linux. It needs its seed that we're hoping to plant with our launch.” 

Koukoumidis and Elachqar are well acquainted with the challenges and inefficiencies of AI development in large-scale organizations, as they are alumni of some of the largest companies in tech: Google Cloud and Apple. During their time at these industry giants, they saw firsthand the limitations of siloed AI development and the inefficiencies that come with it. With only a few thousand researchers at even the biggest AI labs, progress remains slow, fragmented, and insufficient to address critical challenges such as safety, robustness, and multimodal AI development. 

“Even for a company like Apple that has almost infinite resources, comparatively to other companies, we were still working in silos. There was still way too much we couldn’t do ourselves, and it was so inefficient,” Elachqar told AIwire. “This is way too big of a technology to be done in silos. We need to work together. We need to work for universities with some of the brightest talents, students, and professors all across the US, and the world, to build AI in a more transparent, open, robust way because it's going to lead to better outcomes for everybody.” 

A Unified Platform for AI Development

The Oumi platform aims to provide an all-in-one solution for AI developers, enabling them to train, fine-tune, and deploy models efficiently across a wide range of environments. Users can work with models ranging from 10 million to 405 billion parameters, leveraging advanced techniques like SFT, LoRA, QLoRA, and DPO to optimize performance. 

Oumi supports both text and multimodal models, including Llama, Qwen, and Phi, and offers built-in tools for data synthesis and curation using LLM judges to streamline dataset preparation. For deployment, the platform integrates with popular inference engines like vLLM and SGLang while ensuring seamless evaluation through standard AI benchmarks. Oumi was designed for maximum flexibility to allow researchers to run models anywhere, from laptops to supercomputing clusters to cloud platforms like AWS, Azure, and GCP. It also provides integrations with both open models and commercial APIs, including OpenAI, Anthropic, Vertex AI, and Parasail. 

From Universities to National Labs: Oumi’s Growing Network

In addition to their progress and collaboration with major universities, Oumi's founders have ambitious plans to expand their collaborations with national labs and high performance computing centers to streamline AI research at scale. They have already worked with Argonne National Laboratory's Argonne Leadership Computing Facility, where they received a grant to test their platform on the Polaris supercomputer, laying the groundwork to simplify running AI workloads on government-funded clusters. 

Now that Oumi is emerging from stealth, they anticipate a wave of new collaborations from institutions they may not have previously connected with, as word spreads through academic and research communities. 

Investors See Oumi's Possibilities for AI Research

Oumi has secured $10 million in seed funding, launching itself as a Public Benefit Corporation. The Seed funding was led by Venrock and Obvious Ventures with contributions from Plug & Play and Ascend. 

The founders say investors were drawn to its vision of creating an open source AI platform akin to Linux for AI. The founders pitched Oumi as a long-term strategy that would establish a foundational, community-driven AI ecosystem, one that democratizes access to AI research while still allowing for enterprise monetization down the line. 

Their approach mirrors the Red Hat model for Linux, where the core technology remains open and accessible, but specialized enterprise offerings provide a revenue stream. Investors saw the potential in this model, recognizing that by fostering a collaborative AI development environment, Oumi could tap into contributions from the global research community, both in terms of talent and computing resources, without bearing the immense costs of building AI in isolation. 

Elachqar says investor interest in Oumi was strong, allowing the team to be selective in choosing backers who aligned with their mission. The overwhelming demand from academia for an open AI platform played a key role in convincing investors, as universities increasingly feel sidelined in the development of cutting-edge AI research. 

“We were fortunate enough to have an oversubscribed round, so we were able to pick investors that really share our vision,” Elachqar says. “As soon as they talked to any of the professors that we work with, they were all convinced because there is such a strong need from the universities. They all feel so left out, and it's such a pressing problem for them.” 

A Business Model Serving Both Academia and Industry

The name Oumi stands for Open Universal Machine Intelligence, reflecting the company’s core mission. The founders chose “machine intelligence” as a more precise, scientific term for AI, emphasizing their commitment to building truly intelligent systems rather than just following industry buzzwords. The “open” in Oumi underscores their belief that AI should be developed transparently and made accessible to everyone unconditionally. Meanwhile, "universal" represents their vision of AI as a pervasive technology, one that should be available across all platforms, regardless of where it is run or how it is accessed. 

Oumi’s founders chose to establish it as a Public Benefit Corporation (PBC) to balance its mission-driven approach with the ability to operate as a sustainable business. The founders saw three primary options: a traditional for-profit, a nonprofit, or a PBC. While a nonprofit structure might have aligned with their goal of making AI universally accessible, they believed it would be too restrictive, particularly in securing the resources needed to develop advanced AI technologies. At the same time, a standard for-profit model didn’t fully reflect their public benefit mission, which is centered on open, collaborative AI development. The PBC structure provides the best of both worlds: it allows Oumi to attract funding and build enterprise-grade products while ensuring that academic institutions and researchers remain a core part of the ecosystem. The founders also recognized that enterprises tend to have greater confidence in products backed by commercial support, making a PBC structure an ideal way to encourage adoption across both academia and industry. 

A Shared Future for AI Starts Here

The Oumi team sees open source as essential for AI, a field where rapid progress depends on collective innovation rather than siloed, duplicative efforts. If successful, Oumi could reshape the AI research landscape by enabling a global community to work together efficiently, leading to safer, more secure, and more widely beneficial AI advancements. 

The company hopes its open platform will lower the barriers to AI research and development.  As Oumi emerges from stealth, its founders see this as just the beginning of a broader movement that prioritizes transparency, accessibility, and collaboration over competition, paving the way for AI to be developed as a shared global resource. 

“We know that AI is going to be permeating everything that we do, and we want such open technology to be universal, to be accessible to everyone and present everywhere,” Koukoumidis says.