Artificial Intelligence Unlocks New Insights from Wearable Data
By Chibuike K. Uwakwe, MD-PhD Student
Wearables, such as smart watches and smart rings, are typically equipped with a range of sensors that provide information intended to boost users’ understanding of their current physical health, enabling them to maximize their “healthspan” – that is, increasing longevity with good health. Wearables continuously track data, such as heart rate, physical activity, and blood oxygen saturation, which are important metrics for monitoring and maintaining health.
Health risks related to diabetes can significantly reduce healthspan. Continuous glucose monitors (CGMs), another type of wearable, have introduced opportunities to monitor health related to diabetes risk. One reason the commercialization of CGMs has been so impactful is the high prevalence of diabetes and prediabetes in the world. The Centers for Disease Control and Prevention reports that 38.4 million people in the United States have diabetes and another 97.6 million people have prediabetes [1]. The International Diabetes Federation estimates that 589 million people have diabetes worldwide [2].
CGMs enable users with diabetes or prediabetes to track their blood sugar levels in real-time and keep them within a target range. This often involves reacting to spikes and drops in blood sugar levels that may occur throughout the day. The management of blood sugar also plays a role in regulating weight, cholesterol, inflammation, and other aspects of health.
Glucose Responses to Food Vary across Individuals
In January 2017, the Snyder Lab at Stanford University School of Medicine first shared their research on exploring the utility of tracking physiomes and activity using wearable biosensors in gathering health-related information [3]. The group, led by Professor Michael Snyder, the Stanford W. Ascherman Professor of Genetics at Stanford University, found that they could distinguish between insulin-sensitive and insulin-resistant individuals using wearable sensor data, which suggested that wearable sensors could be used to detect risk for type 2 diabetes.
Prof. Snyder and his group at Stanford U. continued their research using CGMs to measure glucose levels in people with type 2 diabetes, prediabetics, and “healthy” people. The Snyder Lab observed that “different people spike to different foods,” whether that be pasta, potatoes, or bread, in a highly personalized manner. This observation sparked the idea of building personalized models of blood sugar response.
Encouraged by the research findings, Prof. Snyder co-founded the company January AI. Ideally, one would be able to anticipate these fluctuations in advance using artificial intelligence.
A Digital Twin to Predict Glucose
The team at January AI developed their own generative AI to predict a person’s blood sugar levels. With wearable platforms generating a large volume of data, machine learning has emerged as a powerful means of processing and interpreting it. January AI’s machine learning models are built upon millions of data points comprising wearable, demographic and user-reported data. In their app, demographic information — including height, weight, age, sex, and health state — can be combined with January AI’s predictive algorithm trained on thousands of users’ real CGM data to create a “digital twin” for precise predictions of blood sugar levels in response to food ingredients. This is based on a food’s glycemic index, a measure of how quickly and how much a food will raise your blood glucose, and the user’s health state (i.e., healthy, prediabetes, or type 2 diabetes).
January AI pairs these predictions with an on-demand AI nutritionist, delivering personalized guidance that evolves with users’ goals, habits, and metabolism. The app enables users to log food easily and accurately with a quick photo, or by describing what they ate in a simple sentence. To be specific, January AI utilizes AI-powered food recognition, allowing users to snap a picture of their meal to instantly analyze its nutritional content and predict glucose impact before they eat. This feature shifts nutrition tracking from passive logging to proactive metabolic optimization.
January AI’s breakthrough technology was not developed in one night. “We’ve spent nearly a decade building January AI so people can see how food affects their blood sugar before they eat it,” said Noosheen Hashemi, Founder and CEO of January AI. “Our goal isn’t to make health complicated; it’s to give people the clarity and confidence to make better choices in real time, without invasive devices.”
AI-enhanced Tools for Behavior Change
Beyond theoretical impacts on user health, there is evidence that AI-enhanced digital health interventions like the January AI app are beneficial. In a study published in npj Digital Medicine, the January AI team found that their app can improve glycemic control and promote weight loss, particularly when users are actively engaged [4].
Hashemi discussed the company’s evidence-based impact saying, “Our research shows that when people are given the right tools — personalized, predictive, and rooted in science — they start changing behavior. That’s what January delivers.”
To Prof. Snyder, these findings are promising because “health coaches don’t scale very well. You can’t get millions and millions of people to have health coaches.” In contrast, the January AI app can solve this scalability issue and help address one of the leading problems impacting healthspan. “The diabetes endemic is more pervasive than the COVID pandemic,” said Prof. Snyder.
Further Health Insights Using AI with Data from Wearables
The January AI app goes beyond just blood sugar management to improve overall metabolic health. The app focuses on setting weekly nutrition targets, based on the user's health goals, via its AI nutritionist, and ingests sleep and activity data from Apple Health Kit. January released a feature that enables users to upload results from recent blood tests to offer more personalized nutritional recommendations and insights, and the team has plans to bring in other kinds of data (such as genomics and microbiome data, etc.) to make its recommendations even more personalized in the near future.
January AI’s use of artificial intelligence to gain new insights from continuous glucose monitoring devices is just one example of how AI can be leveraged to transform wearable data analysis. AI has been used to power the next generation of electronic skin via early disease detection and infectious disease tracing and monitoring [5]. AI agents like Biomni have been shown to perform wearable bioinformatics analyses up to 800 times faster than human experts, presenting a promising solution to managing the mountain of data generated by wearable devices [6]. Another example is the combination of multiple wearable platforms powered by AI, which could create “intelligent healthcare environments” that coordinate to optimize patient safety [7].
With that said, as we embrace the ubiquitous use of AI to unlock insights from wearable data and promote human health, it is important to forefront key issues such as built-in biases and data privacy. By acknowledging the limitations and challenges associated with the use of AI, we can work toward developing more equitable, transparent, and secure wearables that truly improve health outcomes and healthspan for all.
References
[1] Centers for Disease Control and Prevention, National Diabetes Statistics Report: Estimates of Diabetes and Its Burden in the United States, Nov. 29, 2023. [Online]. Available: https://stacks.cdc.gov/view/cdc/148231
[2]Diabetes Atlas, 11th Edition, International Diabetes Federation, https://diabetesatlas.org/resources/idf-diabetes-atlas-2025/
[3] X. Li et al., "Digital health: Tracking physiomes and activity using wearable biosensors reveals useful health-related information," PLoS Biol., vol. 15, no. 1, p. e2001402, 2017.
[4] A. Veluvali et al., "Impact of digital health interventions on glycemic control and weight management," npj Digital Medicine, vol. 8, no. 1, p. 20, Jan. 2025.
[5] C. Xu, S. A. Solomon, and W. Gao, "Artificial intelligence-powered electronic skin," Nature Machine Intelligence, vol. 5, no. 12, pp. 1344–1355, Dec. 2023.
[6] K. Huang et al., “Biomni: A general-purpose biomedical AI agent,” bioRxiv, preprint, doi: 10.1101/2025.05.30.656746, May 2025. Available: https://biomni.stanford.edu/paper.pdf
[7] A. Mahajan, K. Heydari, and D. Powell, “Wearable AI to enhance patient safety and clinical decision-making,” npj Digital Medicine, vol. 8, no. 1, p. 176, 2025.
The eWEAR-TCCI awards for science writing is a project commissioned by the Wearable Electronics Initiative (eWEAR) at Stanford University and made possible by funding through eWEAR industrial affiliates program benefactor Shanda Impact Investment LLC and the Tianqiao and Chrissy Chen Institute (TCCI®).