When we first started looking at nutrition apps, the same issues kept coming up. They stop working the moment you lose signal. The meal suggestions feel copy-pasted regardless of who is using the app. And actually logging what you ate takes way too many steps. So we built NUTS (Nutritional Understanding & Tracking System) — a React Native application that stores all data locally on the device, generates food recommendations based on what each specific user has actually been eating, and gets a meal logged in an average of 3.2 taps. The recommendation engine operates in four stages: nutrient-deficiency detection via rolling averages, cosine-similarity-based food ranking, collaborative filtering from peer consumption histories, and a UCB1 multi-armed bandit algorithm for adaptive personalisation. We ran a two-week study with 72 participants. System Usability Scale score was 76.4. Average feature satisfaction was 4.3 out of 5. This paper walks through how the system is built, how the recommendation engine works, and what we learned from testing it.
We started this project because the apps that were supposed to help people eat better were failing in pretty obvious ways. The three things we focused on — getting it to work offline, making recommendations actually personal, and cutting down how long it takes to log a meal — all turned out to be solvable. Not trivially, but solvable without needing technology that does not exist yet.
The benchmark results came out clearly in NUTS’s favour on offline capability, recommendation quality, and interaction efficiency. The beta study returned a SUS of 76.4 and a mean feature rating of 4.3. Barcode recognition at 85% is the one number we are not happy with yet. Everything else hit the targets we set before building. The modular architecture means adding the features users asked for — social, computer vision, more wearables — does not require rebuilding anything foundational. That work starts now.
References
[1] Aranda-Jan, C. B., Mohutsiwa-Dibe, N., & Loukanova, S. (2014). Systematic review on what works, what does not work and why of implementation of mobile health projects in Africa. BMC Public Health, 14(1), 188.
[2] Burke, L. E., Wang, J., & Sevick, M. A. (2011). Self-monitoring in weight loss: A systematic review of the literature. Journal of the American Dietetic Association, 111(1), 92–102.
[3] Mezgec, S., & Koroušić Seljak, B. (2017). NutriNet: A deep learning food and drink image recognition system for dietary assessment. Nutrients, 9(7), 657.
[4] Nahum-Shani, I., et al. (2018). Just-in-time adaptive interventions (JITAIs) in mobile health: Key components and design principles for ongoing health behaviour support. Annals of Behavioral Medicine, 52(6), 446–462.
[5] Thomas, J. G., et al. (2014). Weight-loss maintenance for 10 years in the National Weight Control Registry. American Journal of Preventive Medicine, 46(1), 17–23.
[6] Trang, N. T., Khanh, P. T., & Nghia, N. D. (2020). Machine learning-based food recommendation systems: A comprehensive review. Journal of Food Science and Technology, 57(9), 3169–3178.
[7] World Health Organization. (2020). Healthy diet fact sheet. Retrieved from https://www.who.int/news-room/fact-sheets/detail/healthy-diet
How to Cite This Paper
Mahesh Kudalkar, Hitarth Shah, Pratham Singh, Rohan Jha (2026). NUTS (Nutritional Understanding & Tracking System): A Smart Mobile Platform for Diet Optimization. International Journal of Computer Techniques, 13(2). ISSN: 2394-2231.
