Reinforcement Learning-Driven Nutrition Coaching
An AI-enabled coach that combines computer vision with reinforcement learning to support diabetes self-management and personalized nutrition.
An AI-enabled coach that combines computer vision with reinforcement learning to support diabetes self-management and personalized nutrition.
Bridging the gap between dietary tracking and adaptive coaching.
Utilizes advanced computer vision to analyze meal photos, estimating calories and macronutrients with high accuracy (over 90%).
An RL agent adapts coaching strategies, meal plans, and nudges based on your unique behavior, preferences, and adherence.
Designed specifically to aid in Type 1 and Type 2 diabetes management through just-in-time adaptive interventions (JITAIs).
The DIEXTRA Framework: Analyze, Adapt, Act.
Take a picture of your meal. Our system uses generative AI and computer vision to identify food items and portion sizes instantly.
Receive real-time estimates for calories, protein, carbs, and fats. The system benchmarks against internal ground truth for precision.
The RL agent updates your meal plans and sends personalized nudges—congratulatory or encouraging—based on your streaks and goals.
Track your progress with a comprehensive dashboard displaying nutrition scores, calorie totals, and adherence streaks.
Based on published research in Reinforcement Learning and Nutrition Coaching.
DIEXTRA operationalizes a JITAI-style loop (Just-in-Time Adaptive Interventions) in two interconnected reinforcement cycles: