Noeri launched on November 16 with $110 million to help businesses tailor AI to their own work, aiming to cut the cost of demanding tasks through specialized models. Menlo Ventures led the round, with Nvidia, Cisco Investments and Snowflake among the backers.
Stanford computer science professors Yejin Choi and Carlos Guestrin founded Noeri and serve as co-CEOs. They are pursuing smaller models that coordinate other models and tools to handle particular problems. In their launch essay, the founders describe the first enterprise and scientific partnerships as proofs of concept.
The software examines how a company’s AI handles a task and proposes revisions, such as changing its instructions, replacing a tool or assigning one step to a different model. Noeri says each proposed change comes with an explanation and requires human approval.
Noeri supports proprietary and open-weight models. According to its launch release, customers own their data and the workflows and models developed for them, with third-party model licenses still applying. The company says it measures quality and cost using customers’ own workloads and metrics.
The idea grew partly from helping a Stanford microbiology and immunology team use OpenAI models more cheaply, SiliconANGLE reported, citing Choi’s interview with Bloomberg. Choi said a small model that learned to use older OpenAI models outperformed GPT-5 on that team’s research projects.