Bold claim: The way we test ingredients is missing the most important signals, and Grainge AI is using machine learning to fix that blind spot.
Grainge AI, a US-based startup founded about a year and a half ago by food chemistry and AI researchers from UC Davis, is developing software to pinpoint the testing protocols that truly matter for ingredient applications. The founders call this a “measurement blind spot” that lingers across the food industry, and their mission is to reveal which data actually drives manufacturing success.
Co-founder and CEO Tarini Naravane explains that you can measure thousands of variables, but only a few will be the ones you actually need. “That’s the problem we solve,” she says. “The big question everyone asks is, ‘what data should I measure?’”
Co-founder and CTO Gabriel Simmons adds that the issue is widespread. Manufacturers often set process parameters based on metrics that are decades old—think protein content in a grain—while more nuanced, modern measurements exist but can be more costly. The natural question arises: is the extra measurement worth the price?
This gap leaves manufacturers with insufficient clarity on which newer metrics justify the investment, leading to wasted lab spending and inconsistent expert guidance, Naravane notes.
Machine learning for measurement selection
Grainge AI’s solution uses ML models to determine which data points are most informative for specific manufacturing problems. Their tools integrate with a customer’s data infrastructure to answer questions about which measurements to use.
“ML models can tell you which data is the most informative,” says Simmons. “They provide a natural, data-driven way to figure out what matters, in a way that would take humans far longer.”
Unlike some recent AI trend applications in food that lean on large language models like ChatGPT, Grainge AI focuses on data-driven learning for its core manufacturing use cases. LLMs can automate certain tasks, but reliability can be a challenge, with failure rates around 15% in some scenarios, according to Simmons.
“We want solutions grounded in data from the actual problem,” he notes. The company does use LLM agents for certain internal tasks, but its core offerings rely on data-driven ML for manufacturing challenges.
Real-time manufacturing focus
Grainge AI concentrates on real-time production issues rather than new product development. Naravane describes situations where manufacturers receive new ingredient batches and must quickly adjust formulations.
“We’re tackling manufacturing efficiency, which means there’s very little time to fix problems in real time,” she explains. “These are truly real-time challenges.”
The startup targets mid-sized co-manufacturers and ingredient suppliers facing formulation hurdles. This includes maintaining consistent product performance or nutrition labels amid ingredient variations and identifying ideal uses for non-traditional ingredients sourced from side streams.
Data collection and partnerships
Grainge AI integrates available customer data and collaborates with contract research organizations that can perform the needed measurements. The company is also building ties with method development laboratories that create quality testing protocols.
Naravane emphasizes that these laboratories aren’t customers but collaborators. “We can anticipate novel use cases and new ingredients,” she says.
Industry perception challenges
Beyond typical startup hurdles, the founders point to misperceptions about AI capabilities as a notable challenge. Many firms dismiss AI applications after limited exposure to language models.
“Skepticism often stems from a narrow view of AI,” Simmons contends. “People come to the table with a 5% understanding of what AI can do, or with ideas about a single form of AI. They may overlook other possibilities that could already solve their problems.”
He adds that chasing the most human-like AI isn’t necessarily the path to business success. “For some problems, you don’t need that level of sophistication.”
Origin and future direction
Naravane’s interest in the field grew from managing German foodservice operations, where she handled large events and faced constant recipe reformulation while minimizing waste. Rather than offering traditional consulting, she pursued a data-driven, technical approach and studied food chemistry at UC Davis. There she met Simmons, who has been working on AI topics since 2018 at the university’s AI Institute for Food Systems.
Currently, the company focuses on texture applications, chosen because incorrect texture can derail manufacturing lines. However, Naravane sees flavor profiling as a future frontier.
Looking ahead, Naravane highlights a broader goal: a deeper understanding of ingredients. “We’re dealing with waste streams, repurposing ingredients, and discovering new sources,” she says. “Fully understanding what an ingredient does is a rich topic that will unlock many new uses.”
Simmons envisions transforming formulation from a reactive chore into a proactive tool. “I’d like the food industry to move away from reactive guesswork when conditions change,” he says. “Formulation should be easy and something you can do confidently on demand, with solutions you know are right.” He aims to achieve this transformation for several major customers within the next one to two years.
Privacy and ethics
Naravane stresses that data privacy remains a priority. “We handle customer data responsibly; privacy and security are non-negotiable for our clients.”
If you’d like to see how AI can reshape ingredient testing and formulation decisions, Grainge AI’s approach offers a compelling framework for turning data into reliable, actionable insights across the supply chain.
