🍇 Nutritional AI Website for Retirement Homes (and anyone else!)
During my time as a research assistant at the University of Waterloo, I worked on a project called NutritionVerse that combined two of my favourite things: software engineering and real-world impact. The goal was to create a nutritional AI website designed to help monitor the dietary habits of seniors in retirement homes. We were featured in The Wall Street Journal!
The project was funded under a government contract, with the intent of improving elderly care through technology. Today, one retirement home is already using it daily to track meals and better support residents’ nutritional needs.
The concept is simple:
- Take a photo of a plate of food.
- The system uses AI to estimate calories, protein, and other nutritional values.
- Care providers gain valuable insights into residents’ diets without needing manual logging.
My role was to build the entire web platform from the ground up—frontend, backend, and database design. I developed the application using Next.js, TypeScript, React, and Firebase, ensuring that it easy to use for elderly and healthcare staff.
I then presented the project at uWaterloo's Computer Vision and Intelligent Systems conference.
As an extension of the nutritional AI website, I developed a smart glasses prototype that integrates directly with the platform. The glasses feature a side-mounted camera and a button for quick meal logging. When the user presses the button, the camera captures an image of the meal and sends it to the website via a Wi-Fi hotspot connection and a custom API endpoint I built.
Within seconds, the system processes the image and displays the estimated nutritional information—calories, protein, carbohydrates, and more—on a small screen facing the wearer. This allows users to get real-time feedback on their meals without needing to manually interact with a phone or computer.
The glasses are fully integrated with the web platform, meaning users can later view their complete meal history, along with nutritional breakdowns, directly on the website.
The photo above shows an early prototype that I presented in class. While the initial version is somewhat bulky, future iterations could be made more compact and user-friendly.
And above here is a picture of the result after eating dumplings.