Nutribetics
Nutribetics is a mobile app developed to assist individuals with Type 2 Diabetes in managing their diet. It uses AI to generate personalised meal recommendations based on user health data. A key feature is a GUI-driven AI agent that can autonomously navigate online supermarkets like Sainsbury’s, select ingredients from the suggested meals, and fill the shopping cart — requiring only final approval from the user.
Personalised Nutrition
Get meal plans tailored to your glucose levels, biometrics, and personal food preferences. Every recipe is selected to help you manage (or prevent) diabetes more effectively.
Dynamic Recipe Filtering
Filters and matches recipes using constraint-based logic — factoring in allergens, dietary goals, macronutrient targets, and medical inputs.
Smart Shopping
Uses our in house GUI AI agent powered by GPT-4o and selenium webdriver to buy your ingredients by autonomously shopping for you.
Idea and Design Stage
This project was part of a 2-day hackathon hosted by The Royal Society of Medicine, themed around digital healthcare. Our core themes were:​
​
-
Briding Gaps between Pricary Care and Community. innovating patient screening and refferal processes to enhance healthcare delivery
-
Improving Access to healthcare: Addressing disparities faced by minoring groups.
​​
We initially explored more niche ideas around the NHS but then thought about combining a GUI with an AI agent to buy ingredients for us. This was something that felt super innovative and unlike anything we’d seen before. Our idea: NUTRIBETICS.
​
Our website/app would collect patient data, process it, and output 10 different uniquely styled recipes from which our customer could choose. Our agent would then process that information and autonomously control the customer's browser to buy the best ingredients.
​
A special thanks to Joshua, Priya and Likitha for their every effort making this project a reality.
​
What makes this project unique!
Multiple "Agents" Working
​​
Most models have a singular LLM call just at one end to execute a very specific task. However here we've got multiple calls all executed without any human intervention. We have the following agents:
-
Nutrition Advisor Agent:
-
Recipe Selection Agent:
-
Ingredient Extractor Agent:
-
Cookie Consent Agent:
-
Web Automation Agent:
​
The Future "Internet of Agents" (Discussion)
​
When the agents evolve, the agents can also talk not only to each other but to external services in a broader network sometimes referred to as an "Internet of Agents." In the latter:
​
-
Individual AI services handle specialized tasks (e.g., grocery automation or personalized nutrition).
-
Individual services can easily share data, constructing ever more specialized tools, which are usable together.
​
Eventually, this allows for customized end-to-end solutions. Imagine connecting a smart fridge inventory monitor, a recipe planner, and an online grocer, all inter-talking without human intervention.
This project shows only one example of how multiple AI-based modules can collaborate to simplify complex tasks that otherwise would be done step by step. As technology continues to improve, you can imagine an ever-expanding "Internet of Agents" where each intelligent service maintains its own niche, all of which feed into automating everything from dinner planning to health care.