JDRF’s Role in Accelerating Artificial Intelligence for Type 1 Diabetes Care

Type 1 diabetes (T1D) management has been transformed over the past decade by continuous glucose monitors (CGMs), insulin pumps, and hybrid closed-loop systems. Yet the next frontier—fully autonomous, AI-driven care—depends on robust data pipelines, predictive algorithms, and large-scale clinical validation. JDRF (formerly the Juvenile Diabetes Research Foundation) has become a central force in this evolution, funding and shaping the artificial intelligence (AI) projects that promise to reduce the daily burden of T1D while improving glycemic outcomes.

JDRF’s strategy is not merely to support isolated AI prototypes; it is to accelerate the entire ecosystem—from data collection and algorithm development to regulatory approval and real-world adoption. By partnering with technology companies, academic labs, and startups, JDRF ensures that AI innovations translate into practical tools for people living with T1D. This article explores how JDRF’s funding, research initiatives, and collaborative networks are driving AI forward in T1D care, and what the future may hold.

JDRF’s Mission and the Case for AI in T1D

JDRF’s stated mission is to accelerate research that cures, prevents, and treats T1D and its complications. AI fits squarely within that mission because T1D generates enormous quantities of data—glucose readings, insulin doses, meal logs, activity levels, and more—that are too complex for any individual or static algorithm to manage optimally. Modern AI, particularly machine learning (ML) and deep learning, can identify hidden patterns, predict future glucose trajectories, and adapt therapy in real time.

“The promise of AI is to give people with T1D more freedom and better outcomes by making the system smarter than any single rule-based program,” explains Dr. Aaron Kowalski, JDRF’s CEO and a longtime advocate for closed-loop technology. “JDRF has been investing in that vision for years.”

Between 2010 and 2024, JDRF committed over $500 million to T1D research, with a growing portion directed toward AI and data science. The foundation’s efforts address three core challenges: data fragmentation across devices, algorithm transparency for clinical trust, and real-world validation of AI-driven interventions.

Why AI Matters for T1D Management

Traditional diabetes management relies on finger-stick blood tests and manual insulin dosing, but even with modern CGMs and pumps, people spend only about 50–70% of time in the target glucose range (70–180 mg/dL). AI can help by:

  • Identifying subtle glucose trend patterns that humans miss.
  • Adjusting insulin delivery proactively before hypo- or hyperglycemia occurs.
  • Personalizing treatment parameters based on individual physiology, activity, and sleep.
  • Reducing the cognitive burden of constant decision-making.

JDRF’s investments target each of these areas, aiming to make AI an invisible but powerful assistant in daily T1D care.

Data Collection and Standardization: The Foundation of AI

Building High-Quality Datasets

AI models are only as good as the data they are trained on. JDRF recognized early that fragmented, non-interoperable device data was a major bottleneck. Through its Data Innovation Fund, JDRF has supported projects that aggregate de-identified CGM, insulin pump, and patient-reported data into large, standardized repositories. One notable example is the Tidepool Big Data Donation Project, which has collected millions of days of real-world diabetes data from volunteers.

By making these datasets available to researchers and developers, JDRF has enabled the training of more robust and generalizable AI models. The foundation also advocates for device manufacturers to adopt common data standards (such as IEEE 11073 and HL7 FHIR) so that AI algorithms can seamlessly ingest information from any compliant CGM or pump.

Data Quality and Labeling

For supervised machine learning, data must be accurately labeled—for example, marking times when a person ate a meal, exercised, or experienced hypoglycemia. JDRF funding has contributed to the development of semi-automated labeling tools that use event detection algorithms to reduce the manual burden on researchers. These tools improve the speed and consistency of training data, leading to more reliable AI models.

Predictive Analytics and Risk Forecasting

One of the most direct applications of AI in T1D is predicting future blood glucose levels. JDRF has supported multiple research groups working on recurrent neural networks (RNNs) and transformer models that learn from sequential CGM data to forecast glucose 15–60 minutes ahead. Accurate prediction is the bedrock of automated insulin delivery (AID) systems, also known as artificial pancreas systems.

Hypoglycemia Prediction Algorithms

JDRF-funded studies have demonstrated that AI can predict impending hypoglycemia with high sensitivity and specificity. For instance, researchers at the University of Virginia, with JDRF support, developed a machine learning model that uses CGM trends, insulin-on-board, and heart rate variability to warn users 30 minutes before a low occurs. This type of predictive alert gives people time to take preventive action, reducing the fear and frequency of severe lows.

Glycemic Variability Scoring

Beyond simple prediction, AI can quantify glycemic variability—a metric linked to long-term complications. JDRF has funded the creation of composite variability scores that combine multiple CGM-derived metrics (standard deviation, MAGE, LBGI, HBGI) into a single interpretable number. Clinicians use these scores to tailor therapy adjustments, and the scores can be fed back into AI models for even better personalization.

Closed-Loop Insulin Systems: JDRF’s Signature AI Achievement

JDRF’s most visible success in AI-driven T1D care is the development of hybrid closed-loop (HCL) insulin pumps, commonly called artificial pancreas systems. These systems use AI algorithms to automatically adjust basal insulin delivery based on real-time CGM readings, while still allowing the user to manually bolus for meals.

From Research to Commercial Systems

JDRF’s first major closed-loop initiative, launched in 2006, brought together engineering teams at the University of California, Santa Barbara, and the University of Virginia. The result was the Zone Model Predictive Control (MPC) algorithm, which became the foundation for several commercial products. In 2017, JDRF partnered with Medtronic to fund the pivotal trial for the MiniMed 670G—the first approved hybrid closed-loop system. Since then, JDRF has continued to support iterative improvements, including the integration of faster-acting insulins and dual-hormone (insulin + glucagon) pumps.

Today, systems like the Tandem t:slim X2 with Control-IQ (which incorporates JDRF-funded research from the University of Virginia) demonstrate the power of AI in practice. Control-IQ uses a predictive algorithm to adjust basal rates and, when needed, deliver automatic correction boluses. JDRF continues to disseminate information about these technologies to help patients and providers understand how AI improves outcomes.

Advancing to Fully Closed-Loop

JDRF’s current goal is to achieve a fully closed-loop system that requires no user input for meals or exercise. This involves advances in AI to estimate carbohydrate content from meal images, detect exercise onset from sensor data, and manage stress-related glucose spikes. JDRF is funding projects that combine computer vision with CGM data to predict meal size and composition, as well as research into reinforcement learning algorithms that can optimize insulin delivery policies in uncertain, real-world conditions.

Personalized Diabetes Management Apps and Digital Coaching

Beyond hardware, JDRF supports AI-powered software that provides individualized recommendations. These apps analyze data from multiple sources—CGM, pump, smartwatch, manual logs—to generate actionable insights.

The Role of Machine Learning in Daily Decision Support

Apps like Glooko and Dexcom Clarity (both of which have benefited from JDRF-funded studies) use ML to generate pattern reports, such as “Your glucose tends to rise steeply after breakfast on weekends” or “You are at higher risk of nighttime lows on days with afternoon exercise.” JDRF has also funded the development of conversational AI (chatbot-like interfaces) that can answer user questions about insulin dosing, sick-day rules, and travel advice in a natural, accessible manner.

Behavioral Nudges and Gamification

JDRF recognizes that technology alone is not enough—user engagement is critical. Some of its funded projects incorporate reinforcement learning agents that learn which types of reminders or encouragements work best for an individual. For example, an AI might learn that a user responds better to a “you’re doing great” message than a clinical alert. This personalized approach can improve adherence to glucose monitoring and reduce burnout.

Innovative Collaborations: JDRF as a Catalyst

JDRF’s impact on AI in T1D care is amplified by its role as a convener and funder of cross-sector collaborations. The foundation has established strategic partnerships with:

  • Dexcom – co-funding research on CGM-based predictive algorithms.
  • Insulet Corporation – supporting development of the Omnipod 5 automated insulin delivery system, which uses an Android-based controller and AI-driven dosing logic.
  • Google – exploring machine learning for glucose prediction and health data interoperability.
  • TypeZero Technologies (now part of Tandem) – commercializing the University of Virginia’s control algorithms.
  • Academic centers like the Barbara Davis Center for Diabetes and Joslin Diabetes Center – conducting clinical trials for AI-enabled interventions.

The JDRF Artificial Pancreas Consortium

Launched in 2015, the consortium brings together more than a dozen research sites to share data, standardize trial protocols, and speed regulatory approval of AI-driven devices. This collaborative structure has reduced the time from algorithm invention to clinical deployment by years. Consortium members have published dozens of studies validating the safety and efficacy of AI-based control algorithms in home settings.

Clinical Impact and Real-World Outcomes

The real-world impact of JDRF’s AI investments is measurable. According to a 2023 meta-analysis of JDRF-funded closed-loop trials, users of HCL systems achieved an average 12–15% increase in time-in-range (TIR) compared to sensor-augmented pump therapy, with a corresponding reduction in hypoglycemia. Moreover, studies show that AI-driven predictive alerts reduce severe hypoglycemic events by up to 50%.

Patient-reported outcomes are equally positive. Survey data collected by JDRF indicate that users of AI-enhanced devices report lower diabetes distress, improved sleep quality, and greater confidence in managing glucose in public or social settings. For parents of children with T1D, the reduced need for overnight monitoring is transformative.

Health Economics of AI in T1D

JDRF has also funded health economic analyses showing that AI-driven systems can be cost-effective over the long term by reducing emergency department visits, hospitalizations for diabetic ketoacidosis (DKA), and long-term complication costs. A 2022 study published in Diabetes Technology & Therapeutics (with JDRF support) estimated that widespread adoption of AI-powered closed-loop therapy could save the U.S. healthcare system $1.5 billion annually by 2030.

Challenges and Ethical Considerations

Despite the progress, JDRF acknowledges several challenges that must be overcome to realize the full potential of AI in T1D care.

Data Privacy and Security

AI models require vast amounts of personal health data. JDRF funds research into federated learning approaches, where algorithms are trained across multiple sites without raw data leaving local servers. The foundation also advocates for strong encryption standards and transparent data-use policies to maintain patient trust.

Algorithm Bias and Generalizability

AI models trained primarily on data from white, affluent populations may perform poorly in diverse groups. JDRF is actively funding projects that collect data from underrepresented populations (including racial/ethnic minorities, low-income individuals, and older adults) to ensure that AI tools work for everyone. The foundation also supports research on fairness-aware machine learning to detect and mitigate bias in glucose prediction models.

Regulatory Hurdles

AI-driven medical devices must undergo rigorous FDA review. JDRF works with regulators to develop adaptive trial designs and real-world evidence frameworks that can speed approval for AI algorithms that improve over time. The foundation also provides educational resources for researchers navigating the regulatory pathway.

Integration with Mental Health and User Experience

AI systems that generate frequent alarms or complex advice can contribute to alert fatigue. JDRF invests in human-centered design research to create interfaces that are intuitive and respectful of user attention. This includes work on adaptive thresholds that reduce false alarms and voice-based interactions that minimize screen time.

Future Directions: What JDRF Is Investing in Next

JDRF’s current research roadmap for AI in T1D includes several ambitious projects:

  • Dual-hormone closed-loop systems that incorporate glucagon or pramlintide to better manage post-meal spikes and exercise-related lows.
  • AI-powered exercise detection and management using wearable sensors (accelerometers, heart rate monitors) to automatically adjust insulin delivery during physical activity.
  • Computer vision apps that estimate carbohydrate content from smartphone photos, integrated into bolus calculators.
  • Predictive models for long-term complications that use CGM and metabolic data to identify individuals at high risk for retinopathy or nephropathy years before clinical onset.
  • AI-based digital twins of individual patients, allowing clinicians to simulate therapy changes in silico before implementing them in the real world.

JDRF is also exploring the potential of large language models (LLMs) to serve as conversational diabetes educators, capable of answering complex questions about insulin dosing, sick-day rules, and travel adjustments with high accuracy. Pilot studies funded by JDRF are evaluating safety and usability.

How the T1D Community Can Get Involved

JDRF encourages people with T1D to contribute to AI research by donating their device data through programs like Tidepool’s Big Data Donation Project. Participation helps researchers train better models while maintaining strict privacy protections. JDRF also runs patient advisory committees that review AI studies, ensuring that the user perspective is integrated from the start.

For researchers and entrepreneurs, JDRF offers various funding mechanisms—from early-stage innovation grants to large-scale consortium awards—specifically focused on AI and data science. The foundation’s research portal details current opportunities and strategic priorities.

Conclusion

JDRF’s contributions to developing artificial intelligence in T1D care are foundational and far-reaching. From catalyzing the first hybrid closed-loop systems to building the data infrastructure needed for next-generation algorithms, the foundation has positioned AI as a critical component of modern diabetes management. While challenges around equity, privacy, and user experience remain, JDRF’s sustained investment in collaborative, ethical, and patient-centered AI research offers a clear path forward. As algorithms become more intelligent and devices more seamless, the goal of a truly autonomous, worry-free life with T1D moves closer to reality.