Noeri Inc., a startup developing artificial intelligence systems that continuously improve their outputs as they learn from an organization’s data and work, is launching with $110 million in funding.
Today’s round was led by Menlo Ventures and saw participation from Seattle-based Madrona Ventures, Kindred Ventures, Wing Venture Capital and NEA. Snowflake and Cisco Ventures also took part in the round.
Noeri is the brainchild of Yejin Choi (pictured, left) and Carlos Guestrin (right), two prominent AI researchers from Stanford University, who spent years developing self-improving systems at the University of Washington prior to moving. The startup’s software is designed to study a company’s data, workflows and business processes.
Then as it learns about them, it will begin to propose changes regarding specific tasks, in order to make them more efficient. The co-founders refer to it as a “living” AI system that gets better at what it does over time.
For instance, it might offer different instructions, suggest different tools or switch out the model that handles each step of a workflow for another. Each of these suggestions comes with an explanation and is sent to a human for review before it’s accepted, similar to how AI-generated code changes are implemented.
Noeri is compatible with both open-weight and proprietary AI models and ensures that customers retain full ownership of their workflows and the AI systems they develop. Choi told Bloomberg in an interview that Noeri is trying to appeal to organizations that find themselves having to choose between either paying fees they can’t afford for the most capable AI models, or else accepting results from a less effective one.
The idea for the company stemmed from an effort to help a microbiology and immunology team at Stanford use OpenAI Group PBC’s models in a more cost effective way. According to Choi, she and Guestrin developed a small AI model that learned how to use OpenAI’s older models, and to their surprise, managed to fine-tune it to perform better than the more advanced GPT-5 on that research team’s projects.
The startup’s name was carefully chosen. Noeri means “intelligence, in the plural,” and it’s meant to reflect Choi’s and Guestrin’s belief that the future of AI won’t be dominated by just one or a handful of extremely powerful AI models.
Guestrin especially is a big proponent of choice in the AI industry. In a speech at Madrona Ventures’ IA40 Summit in September, he notably argued that the future of intelligence cannot be controlled by just one or two companies that own the models.
“Few companies would build their core operations on a single consultant who also advises every competitor, can change its terms at will and can never be brought in-house,” Choi and Guestrin wrote in a blog post outlining the startup’s mission. “A single, centralized AI offers exactly that arrangement.”
Noeri isn’t the only proponent of such a multimodel future. Similar arguments have been made by Mira Murati’s Thinking Machines Labs Inc., which last year debuted Tinker, a cloud-based service that anyone can use to fine-tune AI models and enhance their performance. Similarly, Applied Compute Inc. embeds engineers with its customers to help them fine-tune AI systems.
Guestrin previously helped develop a popular machine learning tool called XGBoost, which is used by researchers to boost the performance of AI models. He also co-founded an early machine learning startup GraphLab Inc., later known as Turi Inc., which was acquired by Apple Inc. in 2016. After that, he joined Apple too and went on to lead that company’s machine learning and AI research organization.
Choi previously led the Allen Institute for AI’s common-sense reasoning research team before joining Nvidia Corp. in 2024 as a distinguished scientist. She quit about a year later to join Stanford in early 2025, from where she helped Guestrin start Noeri in May.
Noeri represents the culmination of two separate research projects. Guestrin pioneered a new self-distillation technique that allows AI models to become their own teachers, turning mistakes into lessons that allow them to improve their responses over time. Meanwhile, Choi’s work on agentic orchestration has shown that smaller models that conduct teams of AI agents and tools can outperform much larger frontier models at lower costs.
Those two projects are the foundation of Noeri’s systems. The startup says it’s now working with partners across multiple industries, including some of its investors, which were deliberately chosen to ensure its systems could be exposed to a wide variety of work.
“Most new AI labs ask the market to wait years for a breakthrough,” said Menlo Ventures Partner Venky Ganesan. “Carlos and Yejin are doing world-class research and putting it to work inside companies at the same time. That combination is rare, and it’s why we moved quickly to lead this round.”
Photo: Noeri