A brand new firm referred to as Oumi has simply been launched, rising from stealth at the moment with $10 million in seed funding. The corporate’s founders describe Oumi because the world’s first unconditionally open supply AI platform, positioning it as a collaborative and clear different to at the moment’s proprietary AI ecosystems, much like how Linux revolutionized working methods.
AIwire spoke with Oumi co-founders Manos Koukoumidis and Oussama Elachqar to search out out what this new platform means for the AI world.
Making AI Analysis Extra Accessible
Koukoumidis, Oumi’s CEO, explains: “What we're saying is the primary unconditionally open AI platform that may make it doable for the tens of hundreds of AI researchers, scientists, and builders around the globe to not simply to extra simply leverage and harness frontier AI, but in addition, extra importantly, to collaborate on the identical platform, working collectively to advance frontier AI in a collective means that may result in extra clear and accountable growth.”
Oumi is developed in collaboration with 13 main AI universities within the U.S. and the U.Ok., bringing collectively among the most revered names in expertise and innovation, together with College of Illinois Urbana-Champaign, Carnegie Mellon College, Princeton College, Caltech, UC Berkeley, GeorgiaTech, Stanford College, College of Washington, New York College, MIT, College of Waterloo, College of Cambridge, and College of Oxford.
The initiative emerged from discussions with high teachers, together with Ruslan Salakhutdinov and Amit Acharya from Carnegie Mellon College, who highlighted the rising challenges AI researchers face. These embody issue conducting large-scale experiments and integrating instruments and infrastructure, even when assets like GPUs can be found.
Koukoumidis says, “Salakhutdinov was telling me on a Sunday night time how onerous it’s for his college students to do the sort of analysis. He mentioned, ‘, the previous few years have develop into so much more durable than it was earlier than. We can not contribute to the frontier anymore. In relation to bigger scale experiments, even smaller scale, it's too onerous for college students to determine all of the various things that have to be put collectively to do the sort of analysis.’”
Oumi co-founders Oussama Elachqar and Manos Koukoumidis. (Supply: Oumi)
The founders then spoke with different high AI researchers and PhD college students who additionally expressed concern about academia's diminishing position in AI innovation.
“What about the way forward for AI, if it's turning into a black field that we simply interface with an API? That's not what academia ought to do. It's not how AI and the sciences ought to be superior,” Koukoumidis relays.
Why AI Wants Its Linux
Oumi goals to be the Linux of AI, offering an open supply different to the dominant, closed AI ecosystems. Not like proprietary AI methods that perform as black containers, Koukoumidis explains that Oumi is constructed round open code, open information, open mannequin weights, and open collaboration, making it simpler for researchers to construct upon one another’s work fairly than fighting fragmented, ad-hoc setups.
The founders additionally consider that AI analysis at the moment is hindered by pointless complexity as high researchers and PhD college students waste worthwhile time cobbling collectively infrastructure fairly than specializing in advancing the science. By providing a unified, clear platform, Oumi seeks to take away these obstacles, very similar to how Linux revolutionized working methods by fostering widespread collaboration and standardization.
Massive Tech Gained’t Dominate Without end
As we noticed this week with the explosive information of DeepSeek and its affect on the AI panorama, the businesses constructing the biggest and costliest AI fashions received’t essentially be on high perpetually.
“I feel they're going to be attempting to play catch up after some time with what's going to be occurring in open supply,” Koukoumidis says. “However to make this occur, open supply wants its Linux. It wants its seed that we're hoping to plant with our launch.”
Koukoumidis and Elachqar are nicely acquainted with the challenges and inefficiencies of AI growth in large-scale organizations, as they’re alumni of among the largest firms in tech: Google Cloud and Apple. Throughout their time at these business giants, they noticed firsthand the constraints of siloed AI growth and the inefficiencies that include it. With only some thousand researchers at even the most important AI labs, progress stays gradual, fragmented, and inadequate to handle crucial challenges corresponding to security, robustness, and multimodal AI growth.
“Even for an organization like Apple that has virtually infinite assets, comparatively to different firms, we had been nonetheless working in silos. There was nonetheless means an excessive amount of we couldn’t do ourselves, and it was so inefficient,” Elachqar informed AIwire. “That is means too large of a expertise to be performed in silos. We have to work collectively. We have to work for universities with among the brightest abilities, college students, and professors all throughout the US, and the world, to construct AI in a extra clear, open, strong means as a result of it's going to result in higher outcomes for everyone.”
A Unified Platform for AI Growth
The Oumi platform goals to offer an all-in-one answer for AI builders, enabling them to coach, fine-tune, and deploy fashions effectively throughout a variety of environments. Customers can work with fashions starting from 10 million to 405 billion parameters, leveraging superior methods like SFT, LoRA, QLoRA, and DPO to optimize efficiency.
Oumi helps each textual content and multimodal fashions, together with Llama, Qwen, and Phi, and gives built-in instruments for information synthesis and curation utilizing LLM judges to streamline dataset preparation. For deployment, the platform integrates with fashionable inference engines like vLLM and SGLang whereas guaranteeing seamless analysis by way of normal AI benchmarks. Oumi was designed for max flexibility to permit researchers to run fashions wherever, from laptops to supercomputing clusters to cloud platforms like AWS, Azure, and GCP. It additionally supplies integrations with each open fashions and industrial APIs, together with OpenAI, Anthropic, Vertex AI, and Parasail.
From Universities to Nationwide Labs: Oumi’s Rising Community
Along with their progress and collaboration with main universities, Oumi's founders have formidable plans to develop their collaborations with nationwide labs and excessive efficiency computing facilities to streamline AI analysis at scale. They’ve already labored with Argonne Nationwide Laboratory's Argonne Management Computing Facility, the place they obtained a grant to check their platform on the Polaris supercomputer, laying the groundwork to simplify working AI workloads on government-funded clusters.
Now that Oumi is rising from stealth, they anticipate a wave of latest collaborations from establishments they might not have beforehand linked with, as phrase spreads by way of educational and analysis communities.
Buyers See Oumi's Potentialities for AI Analysis
Oumi has secured $10 million in seed funding, launching itself as a Public Profit Company. The Seed funding was led by Venrock and Apparent Ventures with contributions from Plug & Play and Ascend.
The founders say buyers had been drawn to its imaginative and prescient of making an open supply AI platform akin to Linux for AI. The founders pitched Oumi as a long-term technique that might set up a foundational, community-driven AI ecosystem, one which democratizes entry to AI analysis whereas nonetheless permitting for enterprise monetization down the road.
Their method mirrors the Purple Hat mannequin for Linux, the place the core expertise stays open and accessible, however specialised enterprise choices present a income stream. Buyers noticed the potential on this mannequin, recognizing that by fostering a collaborative AI growth surroundings, Oumi may faucet into contributions from the worldwide analysis group, each when it comes to expertise and computing assets, with out bearing the immense prices of constructing AI in isolation.
Elachqar says investor curiosity in Oumi was sturdy, permitting the workforce to be selective in selecting backers who aligned with their mission. The overwhelming demand from academia for an open AI platform performed a key position in convincing buyers, as universities more and more really feel sidelined within the growth of cutting-edge AI analysis.
“We had been lucky sufficient to have an oversubscribed spherical, so we had been in a position to choose buyers that actually share our imaginative and prescient,” Elachqar says. “As quickly as they talked to any of the professors that we work with, they had been all satisfied as a result of there may be such a robust want from the colleges. All of them really feel so neglected, and it's such a urgent drawback for them.”
A Enterprise Mannequin Serving Each Academia and Trade
The identify Oumi stands for Open Common Machine Intelligence, reflecting the corporate’s core mission. The founders selected “machine intelligence” as a extra exact, scientific time period for AI, emphasizing their dedication to constructing actually clever methods fairly than simply following business buzzwords. The “open” in Oumi underscores their perception that AI ought to be developed transparently and made accessible to everybody unconditionally. In the meantime, "common" represents their imaginative and prescient of AI as a pervasive expertise, one which ought to be obtainable throughout all platforms, no matter the place it’s run or how it’s accessed.
Oumi’s founders selected to ascertain it as a Public Profit Company (PBC) to stability its mission-driven method with the power to function as a sustainable enterprise. The founders noticed three major choices: a standard for-profit, a nonprofit, or a PBC. Whereas a nonprofit construction might need aligned with their objective of constructing AI universally accessible, they believed it could be too restrictive, notably in securing the assets wanted to develop superior AI applied sciences. On the similar time, a regular for-profit mannequin didn’t absolutely replicate their public profit mission, which is centered on open, collaborative AI growth. The PBC construction supplies the most effective of each worlds: it permits Oumi to draw funding and construct enterprise-grade merchandise whereas guaranteeing that educational establishments and researchers stay a core a part of the ecosystem. The founders additionally acknowledged that enterprises are likely to have higher confidence in merchandise backed by industrial help, making a PBC construction a great technique to encourage adoption throughout each academia and business.
A Shared Future for AI Begins Right here
The Oumi workforce sees open supply as important for AI, a area the place speedy progress is dependent upon collective innovation fairly than siloed, duplicative efforts. If profitable, Oumi may reshape the AI analysis panorama by enabling a worldwide group to work collectively effectively, resulting in safer, safer, and extra extensively helpful AI developments.
The corporate hopes its open platform will decrease the obstacles to AI analysis and growth. As Oumi emerges from stealth, its founders see this as only the start of a broader motion that prioritizes transparency, accessibility, and collaboration over competitors, paving the best way for AI to be developed as a shared international useful resource.
“We all know that AI goes to be permeating every part that we do, and we would like such open expertise to be common, to be accessible to everybody and current in every single place,” Koukoumidis says.
