What makes this more useful than the existing model of scanning a barcode or using a meal tracker or just reading the nutrition facts as you're preparing a meal?
why don't you do this technology without the AI for image recognition and instead keep the AI that predicts levels over time but instead it uses nutrition facts? i'm REALLY unaware of foods that one can buy that don't have even basic facts about what is in it. all legit restaurants have it on their online menus. food you buy in stores have it. even for multi step prepared foods you can just combine the facts for all the parts plus some voodoo with the cooking method to come up with the in vivo response estimation.