Last summer, UK-based personalized nutrition startup ZOE raised a $15mm round from Coefficient Capital to expand its program to the US. Last month, HNGRY finally decided to take the program for a spin, beta tasting its control meal cookies, measuring blood glucose levels, drawing my own blood, sending stool samples, and logging meals over the course of six days. Weeks later, I received an in-depth report containing scores for my blood glucose control, blood fat control, microbiome health, and my overall diet quality during the period. But perhaps the most interesting part of these diagnostics is joining more than 120k participants who have helped train ZOE’s machine learning algorithms to answer the age-old question– “what should I eat?” Thanks to my intensive week of testing and logging, I can now take a photo of any dish and instantly receive a score on the overall health impact of a meal across a multitude of metabolic vectors. Last week’s deep dive featured my experience using the first over-the-counter continuous glucose monitor to get a basic overall sense of my metabolism. Today's piece looks at the richer application layer of applied machine learning on at-home diagnostic testing and food logging to help people make smarter decisions about their diets.