Oumi, the enterprise AI platform that provides enterprises with the ability to automatically develop specialized AI models, today unveiled a set of new capabilities that turn companies into AI factories, completing the entire AI development lifecycle. Enterprises can now build a specialized model, deploy it in one click, and monitor it in production, with every interaction feeding back into retraining, so the model keeps continuously improving on the enterprise’s own production data. To improve AI development flexibility, Oumi is also introducing a publicly available, agent-operable CLI.

This launch arrives at a pivotal moment for enterprises, as the industry shifts from renting intelligence to owning it. Enterprises are realizing that consuming the same general, closed frontier models as their competitors is not a sustainable strategy: there’s no differentiation when everyone is prompting the same model, proprietary business knowledge leaks to frontier labs with every query, costs run 5 to 10x higher than they need to, and enterprises have no control over their most critical technology.

With this launch, any enterprise can run that entire cycle automatically – building specialized intelligence, holistically evaluating, deploying, capturing failures, retraining, reevaluating, and redeploying. Every interaction makes the AI better and more specialized to their use case, as their IP, with full AI sovereignty, reproducibility, and auditability. Enterprises own, inspect, and export the model weights, the training and test data, and even the recipes for how every step was executed, all executable with Oumi’s Apache 2.0 open-source library.

“Every company is becoming an AI company, but nearly all of them are running the exact same closed, generalized models trained on the public web, not on their own workflows, policies, or edge cases,” said Manos Koukoumidis, co-founder and CEO of Oumi. “The next shift in AI will be enterprises becoming compounding intelligence factories—building specialized AI on their own data, owning the resulting intelligence as IP, and making it smarter with every production run.”

With this launch, Oumi provides compounding AI to the enterprise through:

  • Auto-Scaling Deployments: Trained models can now be deployed for production inference with one click, with GPUs that automatically scale based on traffic. At zero traffic, deployments scale to zero, eliminating idle compute costs. At high volume, they scale automatically to meet demand.

  • Continuous Improvement: Enterprises can automatically retrain their specialized models using data captured directly from production traffic. Failure modes are automatically identified, and converted to training signals.

  • Agent-Operable CLI: The new command-line interface gives developers full access to Oumi’s training, evaluation, and deployment workflows directly from the terminal. AI coding agents can operate those same workflows, enabling both human and agent-driven AI development.

Oumi is enabling every company to become a compounding AI factory, and is the first platform to automate the entire loop from model development to deployment and continuous improvement in production. To learn more about how to build, own and compound your own AI, join the Oumi team for a live product demo on August 11th at 10 am PT here.

About Oumi:

Oumi is an enterprise AI platform that lets companies build, own, and continuously improve specialized intelligence – instead of renting general intelligence from frontier model providers. What Claude Code did for software development Oumi does for AI model development: describe what you need and it runs the rest. Oumi closes the entire loop of AI model development combining automation with full transparency, flexibility and AI sovereignty. Oumi was co-founded by a team of former Google Gemini, Apple, and Microsoft AI and engineering leaders. Oumi is based in Seattle and backed by Venrock, Obvious Ventures, Plug & Play, Mozilla Ventures, and Ascend. Learn more at Oumi.ai.

Media gallery

About The Author
Business Wire