Diabetech - Diabetes Tech, Research, News How AI-Powered Food Logging Could Transform Diabetes Management
Aug 4, 2025
Aurelian Briner, CEO and co-founder of the Snack app and AI/computer vision expert, discusses AI-driven food logging for diabetes. He talks about photo and voice meal logging, portion and volumetric estimation, integration with CGMs and wearables, personalized glucose predictions, and ambitions for AI nutrition companions. The conversation covers technical challenges and safety measures for reliable, user-friendly tools.
AI Snips
Chapters
Transcript
Episode notes
Volumetric Depth Solves Portion Size Problem
- Snack's core IP uses volumetric depth estimation to convert plate volume into weight via density rather than relying on flat 2D tagging.
- It leverages device sensors like iPhone LiDAR or multi-camera modes and requires food on a plate or table for accurate estimates.
Host's Morning Breakfast Test Matched His Carb Math
- Justin describes using Snack on a breakfast sandwich where the app correctly identified English muffin, fried egg, and cheese from a photo.
- He also logged a 24oz oat milk coffee and the app estimated exactly 16g carbs, matching his practiced personal dose.
Meal Tips Give Actionable Food Tweaks
- Snack shows contextual meal tips that suggest small changes (e.g., add nuts) to slow glucose rise and improve post-meal glucose.
- These tips will become more personalized as AI learns from user data and preferences.