The Giuseppe Platform
A platform for the plant-forms.
Since its inception, NotCo’s mission is to create tasty, affordable, and sustainable plant-based foods in a better way.
To better serve our mission, we created The Giuseppe Platform — an AI system for the chefs & scientists within NotCo to explore the plant-kingdom, accelerate research, and bring new plant-based products to market at record speed.
What started as an algorithm on a single computer has developed into an AI-enabled Research & Development center at NotCo. We use machine learning and data at the core of our R&D to create the textures, flavors, and functionalities of the NotProducts you love.
Giuseppe Biagio is an interactive system utilized by our Research Culinary team to explore plant-based formulations from a Generative Neural Network.
Chefs can give sensorial reviews and document their creations back into Biagio.
This symbiotic relationship between AI and Chefs has generated a collection of thousands of new plant-based innovations specific to NotCo, which keeps the algorithm learning new things over time.
Pineapples & Cabbage, Spinach & Cacao
The heart of Biagio is the Latent Space Formula Generator, a patented algorithm that generates novel plant-based formulations for a given animal target.
This system generates interesting discoveries that help bring the combinatorial space from an infinite number of possibilities to something more tractable – solutions that can be explored in the experimental kitchen.
One suggestion through Biagio was pineapple and cabbage for plant-based milk, which yielded milky and creamy notes in our culinary experiments. They worked so well, that they are now contributing ingredients in our NotMilk formulation!
Goodbye Excel, Hello Biagio
The new version of Biagio extends the formulation tracking from our experimental kitchen to the rest of the product development cycle.
A typical formulation in NotCo can take several turns and pivots as we identify new ingredients and flavors to optimize not just the taste but the scalability of the product, as well.
Biagio is the central source of truth of all these formulations within NotCo – logging the evolution of a product from inception to industry through a proprietary, secure, and patented interface.
Giuseppe Discovery hosts detailed data and rich analytics of thousands of industrial level plant-based ingredients and animal products.
This includes nutritional, functional, and compositional information that help scientists understand which ingredients will be best utilized for the scalability of a product.
With this proprietary dataset, we are also able to build new algorithms to compliment every step of the product development cycle.
Data You Can Eat
The Ingredient & Product Pipeline serves as the vehicle for new data generation into Discovery.
Any ingredient we might consider for formulation is thoroughly analyzed by scientists at the NotLab and uploaded into the Giuseppe Platform.
Food scientists, located in any NotCo development center around the world, can access and compare the large suite of analytics to tailor-make the perfect formulation.
Turbocharged Food Scientists
We built an ingredient recommendation engine right into Discovery, which allows our scientists to pinpoint the appropriate ingredients to match a target functionality, such as foaming or gelation.
A core piece of the recommendation algorithm leverages our patented method for identifying protein structures through infrared spectroscopy.
The Giuseppe recommendations have especially provided useful in reformulating products due to issues arising from sourcing, availability, or pricing of an existing ingredient.
How do we develop formulations to mimic the animal target? Even more, how do we do it without having to collect thousands of data points on each formulation?
Enter Giuseppe Toolbox.
Toolbox allows and chef or scientist to create and optimize over different formulations to achieve a plant-based prototype that hits the texture and functionality of an animal product.
And since the behind-the-scenes AI can adapt to any process and optimize over many parameters simultaneously, it speeds up the initial product development time up to 10x. Toolbox is now being used in most of the new product development within NotCo.
Milk Foaming Validation — A Case Study
The human mind is exceptionally bad with big numbers: when there are many parameters to vary in an experiment, scientists have traditionally resorted to trial and error for figuring out what parameters cause what response.
Instead of trial and error, Toolbox uses AI to directly optimize properties of formulation. As a validation study, scientists decided to see whether Toolbox could come up with a plant-based milk formulation that could optimize its foaming capacity.
The algorithm starts with a blank-slate, knowing nothing about how ingredients relate to milk-foaming. Initially, it explores a few different combinations of ingredients, then iteratively optimizes through feedback, and ultimately converges to an optimal formulation that foams rather well.
This direct optimization of the formulation allows scientists to reduce the number of lab trials needed to get to a desired result – and thus speeding up the development of NotProduct formulations!
Experiments with a Side of Analytics
Sometimes Toolbox recommends formulations that are non-intuitive, yet work well in optimizing our plant-based prototype. We call these “wow factors”.
To further assist scientists, Toolbox comes prepackaged with a suite of charting tools to help explain the “Why” behind the AI-constructed formulation.
Did you know that flavor is sensed primarily through your nose rather than your mouth?
The flavors you taste when taking a bite come primarily from aroma compounds that exist naturally in all foods.
We developed Flora as part of the Giuseppe Platform in order to understand the mapping of aroma compounds across the animal and plant-worlds.
With over 30,000 molecules ingested into the system, Flora recommends aroma compounds from plants that can be recombined to mimic the taste of dairy or meat in our NotProducts.
Going down to the atomic level
When you take the oils and extracts of plants, you are distilling the essence of the naturally occurring compounds that occur in that plant.
Flora helps identify and surface up these plants that can used for oils and extracts in the development of complex flavors.
Hm, Tastes like Chicken
For a real example of the utility of Flora, take a look at NotChicken, which was recently launched in Latin America.
Flora helped identify oils and extracts from tomatoes, strawberries, and peaches that could be used to develop complex flavors that – unbeknownst to the flavor scientist – actually tastes like chicken.
This system is powered by scientists trying interesting and novel suggestions from our core algorithms. With R&D backed by AI, exploring novel solutions isn’t as time-consuming as it used to be and can yield tremendous results.