Monday, December 23, 2024

Climax Foods Gains Competitive Edge in Sustainable Foods Market through Machine Learning with Benchling

Company gets products to market faster in crowded next-gen foods market through innovative, data-driven approach to R&D.

Climax Foods, Inc., a food-tech company using data science and machine intelligence to create animal-free products from plants, has gained a competitive edge in the next-generation foods space by embracing data science and relying on Benchling as its central source of truth for R&D data, collaboration, and insights.

Growing demand for environmentally-sustainable foods and fast-evolving consumer preferences have disrupted the traditional consumer packaged goods (CPG) and agriculture industries. Sales of plant-based dairy and meat alternatives alone are expected to increase 5x from 2020 to 2030, reaching $162 billion. Founded by data scientists formerly of Google, Impossible Foods, and SpaceX, Climax Foods has differentiated in this competitive next-generation foods market with an approach to cheese that is all about data, not dairy. Every property of the company’s plant-based cheeses provides a measurable data point — flavor, texture, meltability, color, nutritional value, scalability, cost, and sustainability — to be modeled and prototyped in an assay-model-feedback loop that applies machine learning (ML) to continuously improve and develop new digital recipes. Benchling has been key to enabling this, providing the cloud-based platform that leads to FAIR (findable, accessible, interoperable, reusable) data and that supports ML including featurization, data correctness and quality checks, and model training.

Compared to conventional trial-and-error methodology, Climax Foods’ innovative R&D yields faster cycle times and superior end products. Climax Foods has been able to master difficult-to-produce artisanal cheeses like brie, feta, chevre, and blue cheese in just two to three years. The company also recently announced the discovery of the first-ever plant-based protein ingredient replicating the melt and stretch of the dairy protein casein using ML, and credits the Benchling platform as critical to this work and innovation. With its strong product momentum, Climax secured a partnership to co-develop new products with Babybel cheesemaker The Bel Group, the world leader in branded cheese and a major global food player.

“Sustainable foods are the future and data science is how we achieve this,” said Karthik Sekar, PhD, Head of Data Science at Climax Foods. “Our building blocks are the three hundred thousand varieties of plants, which can be combined in millions of different ways to reach specific textures, flavors, smells, and environmental impact. This is a huge combinatorial screening problem that even the largest labs can’t crack when done manually. Climax Food is uniquely positioned to meet this challenge with our focus on data and machine learning.”

Also Read: Olo Launches Card-Present Payment Processing via Kiosks to optimize in-store ordering experience and simplify payments management

Climax Foods began working with Benchling in 2021 as the next-generation foods company was going through an inflection point in their growth, expanding team, product line, and customer base. The company knew that its use of ad-hoc spreadsheets for data entry and tracking was untenable, a barrier to scalability, and a poor use of scientists’ time. Climax Foods had previously considered building its own databases and query tools, but recognized that would be too expensive, time-consuming, and potentially inferior to a third-party solution.

With Benchling, Climax Foods focused on three key areas to enable its data and ML-driven operations: data management and centrality; adaptability and scalability; and consistent, cross-team ontology. The status quo had been for data to exist separately across different files and Google Sheets. When running queries, individuals had to track down each table, set up joins, and retrieve results. The prototyping team was outgrowing the limitations of Google Sheets, which could not accommodate multiple users inputting high volumes of data simultaneously. The company couldn’t rely on data engineers to input and model new data and workflows, an approach that was both unwieldy and costly.

Now, with Benchling’s centralized platform, all data resides in a single place, where queries are seamlessly executed. This centralized approach improves data integrity, preventing data from being lost, incomplete, or mischaracterized. Climax Foods generates new experiments and protocols weekly, if not daily, and Benchling’s highly modular and adaptable platform enables this speed and velocity of science.

Climax Foods also needed to manage more sophisticated data science, optimizations, and ML as it evolved its R&D. For example, scientists’ models needed to track variables such as increases in meltability and cheddar flavor without increasing the cost by more than 10%. Having a consistent ontology across teams was critical to these efforts, whereby a classification would mean the same thing across the prototyping and textual contexts. Ontology was nearly impossible to manage in the company’s previous disparate systems. Benchling’s centralized platform organized concepts such as ingredients, formulations, and assays. This has been essential, underpinning experimental efforts so that they are more comparable. Benchling came with an excellent ontology out-of-the-box that the Climax team could build upon.

Working with Benchling, Climax Foods also accelerated its ML efforts. “There’s a lot of excitement for how machine learning and artificial intelligence can be used in the food industry. At Climax, we’re delivering on that promise today,” said Sekar. “We continuously train our portfolio of machine intelligence tools to enable the plant-based recreation of any taste and texture, while optimizing for nutrition and lowering costs. Benchling specifically gives us the solid foundation where data is findable, accessible, interoperable, and reusable with FAIR data principles. This quality data and Benchling’s flexible system provides the ideal setup for developing and training our machine learning models.”

SOURCE: PRNewswire

Subscribe Now

    Hot Topics