As the CTO of a startup at the intersection of food, technology, and sustainability, I’m often struck by how limited access to climate data remains. For over two decades, there’s been a massive effort to expand knowledge about carbon emissions within the Life Cycle Assessment (LCA) field. Yet, despite progress in translating scientific data into industries like food and beverage, a fundamental challenge persists: existing infrastructure and mindsets are not ready to fully embrace this movement.
When we founded Niatsu, I made a pledge to bridge this gap between data and action. Often, the most pragmatic solutions prove to be the most transformative. This is precisely where Large Language Models (LLMs) like ChatGPT come into play. In sustainability, these technologies should be seen as the long-missing connectors. While consultants spend weeks compiling Excel sheets that quickly become outdated, LLMs – despite their energy consumption – offer the ability to build bridges faster and at scale, unlocking unprecedented potential for real-time insights and action.
Many companies in the food and beverage industry have already encountered some form of carbon accounting, primarily focusing on Scope 1 and 2 emissions. Even if they aren’t familiar with terms like carbon equivalents, they’re no strangers to sustainability labels such as Bio Suisse, Fairtrade, or Demeter. When we engage with customers, there’s a clear tendency for sustainability as a topic, but most companies are still grappling with the fundamentals . Which frameworks should be used? What matters in which country? Haven’t we done in enough my moving towards renewable energy? In fact, many sustainability teams have only been established within the past years.
So, what’s the biggest challenge they face today?
The answer is clear: data. Regardless of where a company is on its sustainability journey, the data landscape is often bleak. Even compiling the basic data for Scope 1 and 2 emissions can be an exhausting, months-long process. And while companies may produce annual sustainability reports, these efforts rarely deliver actionable value. What about using the data to drive decisions? What about addressing the future risks posed by climate change?
Far too much time is spent gathering information from various departments, only to document it in reports that are often outdated the moment they’re published. We’ve spoken to major F&B companies still finalizing sustainability reports from three years ago. From these conversations, one conclusion is clear: the future of effective climate strategies lies in shifting from slow, cumbersome data collection to generating actionable insights – quickly and at scale.
When we developed our database, we realized that the challenge we needed to solve as a data provider went beyond simply having good data. A significant part of the solution lies in building layers that make this data as accessible as possible. By reducing the cost of accessing data and increasing its usability, we aim to generate a higher impact in the long term. As a tech-savvy team, we wanted an engineering challenge.
A key focus in our approach was understanding how to handle two critical aspects of food data: identifying the relationships between different food items and interpreting unstructured information. For instance, determining whether tomato paste is more closely related to fresh tomatoes or tomato puree is not an intuitive tasks. To address this, we trained neural networks capable of understanding ingredients and processed products, and matching them accordingly. Below, you can see how our neural network (with dimensions reduced to 2D for illustration) interprets a range of food items. This model enables us to analyze entire ingredient lists and match them accurately with our database, eliminating the manual process of searching for the right dataset.
Interpreting unstructured information, such as technical data sheets or recipes, presentes an entirely different challenge. The lack of structured data in the food industry is staggering, yet this is precisely where LLMs shine. These models have the potential to transform the way sustainability is approached in the coming years. For instance, our recipe analyzer leverages LLMs to extract ingredients and their quantities directly from recipes. With this data, we can immediately calculate a carbon footprint, generating actionable insights and value for customers almost instantly. You can try carbon chef for free here: recipe.niatsu.com
Having worked at this intersection, I see immense potential for applying this technology across various industries. What excites me the most is its ability to free up time for employees, shifting their focus away from the repetitive accounting and reporting tasks to actually implementing strategies and driving impactful actions. At Niatsu, we are actively experimenting with automated reports that aim to minimize bureaucratic hassle further, enabling teams to spend their time where it truly matters.
And what about the data side of things? To me, any assessment boils down to recursive data-matching processes: whether it’s analyzing life cycle assessment data at the farm level, determining the steps needed to produce a specific product, or crafting visually engaging graphs for sustainability reports. Efficiency and scalability will reshape the way we approach sustainability in the coming years. This shift should be embraced, as it allows key decision-makers to focus on the actions that will truly help mitigate human-induced climate change and adapt more quickly.
NIATSU GMBH IS A COMPANY REGISTERED IN SWITZERLAND (ID: CHE-326.229.899)
© 2025 NIATSU GMBH. ALL RIGHTS RESERVED. OUR
TERMS OF SERVICE APPLY.
