Menlo Ventures has led a $110 million round in Noeri, a five-month-old AI startup co-founded by Stanford professors Yejin Choi and Carlos Guestrin, Bloomberg reported on 16 November. Nvidia, Snowflake and Cisco also invested, along with Madrona Venture Group, Kindred Ventures and NEA, Bloomberg reported. The valuation was not disclosed.
Menlo’s own 2026 enterprise AI report lists Noeri as a Menlo portfolio company and puts specialists like it at 30% of the 10% of AI builders who post-train their own models. That is about 3% of AI builders, by our arithmetic.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer on why products built on the same model differ. It teaches the general idea only and does not discuss Noeri or its round.
The key insight: As we read the figures, the round prices a small market at its start: Menlo’s own survey puts specialists like Noeri at about 3% of AI builders, by our arithmetic, while post-training adoption rose from 6% to 10% in a year, with Nvidia, Snowflake and Cisco among the backers Bloomberg named.
What Bloomberg Reported
Bloomberg’s Dina Bass reported that Noeri is developing “a so-called AI engineer designed to review and suggest improvements for a client’s artificial intelligence models, coding tools and other software”. The recommended changes are explained and sent to a human to approve, the report says.
The goal, per Bloomberg, is to help companies get “the best performing AI systems based on their needs and what they can afford”. Bloomberg named seven investors: Menlo Ventures as lead, Madrona Venture Group, Kindred Ventures, NEA, Nvidia, Snowflake and Cisco. It did not give the valuation or the size of any single investor’s check in the portion of the article we could read.
Choi told Bloomberg the idea came from Stanford research: “We made this small AI that learns to use other AI, including GPT-4, but not 5 because at Stanford we are too poor to use GPT-5 all the time for research”. Bloomberg reported that the tool refined the older models enough to outperform GPT-5 for the team’s particular scientific research needs.

Who Founded Noeri
Bloomberg reported that Choi was awarded a MacArthur Foundation “genius grant” and was previously a distinguished scientist at Nvidia, where Choi developed a small model that learned to use more powerful AI systems to boost performance and reduce costs. Noeri’s home page says Choi is a Professor of Computer Science and Senior Fellow at HAI at Stanford and previously led research at Ai2 and NVIDIA.
Guestrin is director of the Stanford AI Lab, Bloomberg reported. Noeri’s page says Guestrin co-created XGBoost, co-founded Turi, which Apple acquired in 2016, then led machine learning and AI at Apple, and that Noeri is Guestrin’s seventh startup. The page also lists Prithviraj Ammanabrolu, a professor at UCSD, in the founding team, and offices in Palo Alto and Zurich.
What Noeri Says It Builds
Noeri’s site says companies use it to build “specialized AIs from its own data and know-how”. It lists four product steps: the AI “Learns on the job”, “Builds its own tools”, “Trains its own models” and “Tracks your gains”, with quality and cost “measured on your own traffic and metrics, not a public benchmark”.
Its vision page makes the cost case directly: “A model sized to the problem, trained with the right craft, can often match a giant one at a small fraction of the cost and energy”. The page says the company’s research on agentic orchestration showed that “a small model conducting a team of other models and tools can outperform frontier models at a fraction of the cost”.
How Big the Market Is in Menlo’s Own Data
Menlo’s 2026 State of Enterprise AI report says post-training adoption “rose from 6% of AI builders in 2025 to 10% this year, but it remains limited”. Among builders that post-train, it says 91% use outside partners: 50% use infrastructure providers like Fireworks and Modal, 41% use expert-data and evaluation platforms like Mercor, and 30% use specialists like Applied Compute, Trajectory and Noeri, which the report marks as a Menlo portfolio company.
Across all AI builders, that is 9.1% using outside post-training partners, 5% using infrastructure providers, 4.1% using data and evaluation platforms and 3% using specialists, by our arithmetic. Adoption rose about 1.7 times in a year. The report draws on a survey of nearly 400 U.S. enterprise decision-makers fielded by Morning Consult in September, and we covered its post-training figures in our earlier piece.
Menlo names two obstacles: “specialized models need constant updates to keep pace with a frontier that moves every few months, and most application companies lack the expertise to train them effectively”. It says that “creates an opportunity for infrastructure and services providers”.

A Second $110 Million Round With Nvidia and Snowflake
The round matches one from July 2025. Reka, whose release says it is known for its “ultra-efficient multimodal models”, announced a $110 million investment backed by investors “including NVIDIA and Snowflake”. The two rounds are the same size and 16 months apart, by our arithmetic. Reka’s release says its focus on efficient training and serving infrastructure let it develop market-leading models “at a fraction of the cost”; Noeri’s vision page says a right-sized model can match a giant one “at a small fraction of the cost and energy”.
The Structural Read
As we read Noeri’s pages, the product is the layer around the model: it says its AI learns on the job, builds its own tools and trains its own models, with quality and cost measured on the customer’s own traffic rather than a public benchmark.
As we read Menlo’s report, the bet answers the two obstacles it names: specialized models that need constant updates, and application companies that lack the expertise to train them. The report marks Noeri as a Menlo portfolio company, and Menlo led the round, per Bloomberg.
Noeri, vision page
“For many problems of real economic value, the right-sized model can be much smaller than we assume.”
Three Implications
FOR AI BUYERS Noeri says quality and cost are measured on the customer’s own traffic and metrics; it publishes no customers, revenue or pricing on its site.
FOR THE POST-TRAINING MARKET Menlo’s survey puts outside post-training partners at about 9.1% of AI builders, and specialists at about 3%, by our arithmetic.
RELATED Our piece on Menlo’s post-training figures covers the 10% adoption rate and the Cursor and Harvey examples.
The Business Engineer Lens
This story maps onto the Business Engineer framework The State of The Inference Economy.
The framework argues: “Inference accounts for 80 to 90 percent of the lifetime cost of a production AI system.”
As we read it, Noeri is pitched at that running cost: its vision page says a model sized to the problem can match a giant one at a small fraction of the cost and energy. The framework predates the round and does not discuss Noeri.
What Is Not Established
Bloomberg reported that the valuation was not disclosed, and the portion of its article we could read does not give how much Nvidia, Snowflake or Cisco put in. The rest of the article is behind Bloomberg’s paywall and we did not read it. Noeri’s site did not carry an announcement of the round when we checked on 16 November.
Noeri publishes no customers, revenue or pricing on its site. The claim that its tool outperformed GPT-5 for a Stanford research team is Choi’s, as reported by Bloomberg. We did not contact Noeri, Menlo Ventures or Nvidia for this piece.
The Bottom Line
Bloomberg reported a $110 million round in Noeri led by Menlo Ventures, with Nvidia, Snowflake and Cisco among seven named investors and no disclosed valuation. Menlo’s own report puts specialists like Noeri at about 3% of AI builders, by our arithmetic, in a post-training market whose adoption rose from 6% to 10% in a year.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. The round size, lead investor, the six other named investors, the undisclosed valuation, the product description, the founders’ backgrounds and Choi’s quote are from Bloomberg’s 16 November report by Dina Bass; we read the portion available without a subscription, and the rest of the article is behind Bloomberg’s paywall. Noeri’s product steps, team details and cost claims are from its home and vision pages.
The post-training shares are from Menlo Ventures’ 2026 State of Enterprise AI report; Menlo led the round, per Bloomberg. Reka’s figures are from its July 2025 release. The shares of all builders, the 1.7x ratio, the 16-month gap and the investor count are our arithmetic. Noeri, Menlo Ventures and Nvidia were not contacted for this piece.
Sources: Bloomberg, 16 Nov 2026 · Noeri home page · Noeri vision page · Menlo Ventures, 2026: The State of Enterprise AI · Reka release, 22 Jul 2025









