Spotlight on: Konstantinos Andrikos on AI for Smarter Food Choices
Improving food systems requires smart digital solutions that support healthier food choices and more sustainable practices. The WiseFood project stands out for its focus on developing applications that empower citizens to adopt healthier and eco-friendly eating habits.
Konstantinos Andrikos, R&D Engineer and part of the WiseFood consortium, is responsible for the architecture, implementation, integration, and maintenance of the WiseFood platform. He brings valuable experience from Infili Technologies SA, where distributed AI, big data analytics, and robust software solutions are part of daily work.
Given the strong connection between WiseFood and STELAR – both working to strengthen data-driven decision-making in agrifood – we spoke with Konstantinos to hear his perspective. His work with machine learning and information intelligence offers practical insights for improving agrifood data.
How can we make digital solutions truly work for food systems? Read on to explore his thoughts.
Exploring Our Guest's Motivation and Background
Could you share a bit about your professional journey and what drew you to the fields of AI and data science?
I have primarily been involved in research projects leveraging artificial intelligence across a wide variety of application domains. What really drew me to AI and data science is their transformative potential. It is fascinating how these technologies are fundamentally game-changing, offering innovative solutions to challenging problems.
What inspired your interest in improving decision-making around food through technological innovation?
My interest comes from observing some significant challenges we face with our food systems and habits today. We are seeing rising rates of obesity and at the same time an alarming increase in food waste. A major contributing factor is the increasingly fast-paced nature of modern life that often leaves households with very little time for meal preparation, which in turn contributes to a greater reliance on fast food. Added to this complexity is the diverse range of nutritional preferences, dietary needs, and cultural considerations we all have. This situation highlights the need for tools that empower citizens to make better decisions about their food and meal habits, for the good of both themselves and the environment.
The WiseFood Project: Responsibilities and Technical Insights
In your role as R&D Engineer and WP1 leader in the WiseFood project, what are your main responsibilities and areas of focus?
WP1 focuses on the design and implementation of the WiseFood platform. We are targeting to develop a robust, scalable and modular architecture that can effectively support all the smart applications and tools that will be developed throughout the project. We are also responsible for the technical coordination to bring this platform from concept to operational status.
This involves defining technical requirements and managing the integration of all components. Part of our work is overseeing the data infrastructure, which includes identifying, cataloging and preparing diverse data sources such as scientific articles, recipe collections, and nutritional databases.
A crucial aspect of our work in WP1 is integrating advanced Artificial Intelligence, particularly Large Language Models. We are carefully selecting and tailoring these AI models specifically for the food domain in order to empower the WiseFood platform with smart capabilities allowing the different applications to offer personalised advice, understand user queries in natural language and process complex food-related information more effectively.
What specific technical expertise or approaches does Infili Technologies bring to the WiseFood project, and how are these applied in real-world scenarios?
We specialise in designing and building robust software architectures primarily centered around the development of AI-driven platforms. Our track record includes implementing AI systems applied to a wide range of applications, spanning from intelligent automation to data-driven decision-making.
This is powered by our expertise in core AI technologies such as Computer Vision, Natural Language Processing and advanced machine learning, alongside a strong capability in designing intuitive user interfaces to ensure these complex systems are accessible and effective for end-users.
Harnessing AI for Smarter Food Choices
Both WiseFood and STELAR make use of Knowledge Graphs and Large Language Models. How do you see these tools supporting smarter food systems, and what potential do you see in applying similar methods across different domains?
Knowledge Graphs and Large Language Models are a powerful combination for creating smarter food systems. Knowledge Graphs act like intelligent, interconnected maps of food-related information. They do not just store data; they understand the relationships between different pieces of information – for instance, how an ingredient relates to a nutrient, a recipe, a dietary restriction, or its environmental impact. In WiseFood, this allows us to build a rich and reliable foundation of food knowledge that our AI tools can draw upon.
Large Language Models bring the ability to understand and communicate in human language. In WiseFood we are tailoring these LLMs to understand the user’s needs around food and nutrition discussions. This means users can ask questions naturally, as if talking to an expert, and get clear understandable answers. LLMs also help us process and make sense of large volumes of textual information, like scientific articles or dietary guidelines, to build and update our knowledge base more efficiently.
Together, these tools create a smarter food system by making complex information accessible and actionable. Knowledge Graphs provide the verified, structured backbone of facts, while LLMs offer an intuitive way for people to interact with that knowledge and get personalised advice. This synergy can help overcome information overload and misinformation, guiding citizens towards healthier and more sustainable food choices.
The potential for applying Knowledge Graphs and Large Language Models together extends far beyond food systems. Essentially any field that deals with large amounts of complex interconnected information can benefit from this synergy.
In your view, what are the biggest challenges when it comes to implementing AI technologies in areas like food systems and consumer decision-making?
The quality of data is one of the first issues that we face when dealing with AI systems. Especially in food-related domains, we need big amounts of accurate, up-to-date and context-specific data. These data could be detailed nutritional values, local availability of ingredients, sustainability metrics and even cultural food preferences. Gathering and maintaining such comprehensive and reliable data is a continuous effort.
Apart from that, building trustworthy and explainable AI systems is paramount. When AI offers advice about what to eat, people should know why this decision has been made. “Black box” AI, where reasoning is not clear, will not inspire confidence. We need to design systems that are transparent and can explain their outputs in a way non-experts can understand.
Another challenge we often face is achieving true personalisation. Food choices are deeply personal, influenced by health conditions, allergies, budgets, time constraints, cultural backgrounds and taste. Developing AI that can address this complexity for a diverse population is a significant technical and design challenge.
A major concern we are also addressing through this project, is related to the ethical concerns, including potential biases. AI models can often learn and propagate bias that is present in the data it is trained on. This could lead to unfair or even harmful recommendations for certain groups. Keeping the AI aligned with evolving scientific understanding and national dietary policies is also something that is being addressed in the project.
Finally, there is the challenge of user adoption and digital literacy. The most sophisticated AI tool is useless if people do not find it easy to use or do not integrate it into their daily lives. We need to ensure our tools are intuitive and accessible to everyone.
Building Trustworthy, User-Centric AI Solutions
How do you ensure that AI-powered tools are based on solid, reliable science and data, but are also simple enough for non-experts to use and trust?
Ensuring our AI tools are both scientifically sound and user-friendly is a cornerstone of the WiseFood project. To ground our tools in reliable science and data, we are committed to using high-quality and authoritative information sources, such as peer-reviewed scholarly articles, official national dietary guidelines and policies, as well as verified nutritional databases.
We are also dedicated to making the AI explainable, meaning that the system will not just give an answer, but it will be designed to provide reasons for its suggestions in an easy-to-understand way. Transparency is crucial; we intend to disclose to users the sources of the information or the underlying methodology for any calculations presented.
Recognising that AI alone has limitations, we also involve human experts to create a more robust and reliable system. Nutritionists and food scientists collaborate closely with our AI development. Their expertise is vital in the initial stages of data curation, where they help identify, clean and structure the information our AI systems need. Subsequently and just as importantly, they are involved in validating the outputs, ensuring that the information the AI systems use and the advice they provide are not only algorithmically sound but also scientifically accurate, ethically responsible and practically applicable in real-world dietary contexts.
To make these tools simple enough for non-experts to use and trust, we are focusing heavily on user-centric design. This starts with creating intuitive interfaces, particularly leveraging natural language and conversational AI, so users can interact with the tools as if they are having a simple conversation.
A crucial part of achieving this user-friendliness and trustworthiness is our multi-actor approach, which involves establishing Living Labs in different countries. Through these labs, we will co-create and test the applications with everyday citizens, ensuring the tools are practical, meet real needs, and are presented in a way that makes sense to them. The goal is to provide recommendations that are not only accurate but also genuinely actionable and easy to adopt into daily life.
Looking ahead, how do you envision AI contributing to more sustainable and health-conscious food systems in the near future?
I envision AI playing a transformative role in making our food systems much more sustainable and health-conscious in several ways. Firstly, AI will offer highly personalised guidance on diet and nutrition at a massive scale. Imagine an AI assistant that understands your unique health needs, preferences, budget, cultural background, and even what is locally and seasonally available, helping you plan meals that are both delicious and truly good for you and the planet.
Secondly, AI will be instrumental in optimising food supply chains to significantly reduce waste. It can improve demand forecasting, manage inventory more efficiently and help connect producers more directly with consumers, especially for short food supply chains. This means less food thrown away from farm to fork.
Furthermore, AI can inspire culinary innovation towards healthier and more sustainable recipes. It can suggest smart ingredient substitutions or modify existing recipes to boost their nutritional profile and lower their environmental footprint, all while keeping them tasty.
AI will empower both consumers and producers. It can help consumers make more informed, conscious choices and support producers, especially small-scale and local ones, in adopting more sustainable practices and reaching consumers who value them. The overarching goal is a food system that is smarter, fairer, and better for both people and the planet.
Conclusion
Konstantinos Andrikos’s insights highlight the potential of AI to create smarter, more sustainable food systems by combining technological precision with user-friendly design. His work emphasises the importance of trustworthy, transparent, and personalised tools that can truly make a difference in everyday food choices.
Follow our Blog and LinkedIn page for upcoming interviews and the latest updates!
