Traditional nutrition apps require users to search databases, select portion sizes, and manually enter every item in every meal. For a single lunch, that might mean five separate entries with estimated weights and brand selections. Multiply that by three meals plus snacks, and you're spending significant time on data entry instead of actually eating.

The people who need tracking most—those managing diabetes, recovering from eating disorders, or following strict dietary protocols—are often the same people for whom the cognitive load of manual logging creates the most friction. A simpler input method doesn't just save time; it makes consistent tracking actually sustainable.

The app reduces food logging to a single action: take a photo. The AI analyzes the image, identifies the food items, estimates portions, and outputs calorie and macro data. Users get immediate feedback without searching databases or guessing serving sizes.

I built the processing pipeline using n8n for automation, connecting the photo input to an LLM for image analysis and nutrition estimation. The interface was designed in Figma and built with Lovable, prioritizing speed and clarity over feature bloat.

Friends and family demoed the working prototype, and I ran accessibility scans against the screens: 44pt minimum tap targets, contrast checks, text labels on every chip and badge so nothing depends on color alone, and a large-text toggle that scales the base font.

Accuracy was the ceiling, and the ceiling was cost: the cheap models I could afford to test on guessed portions badly, and the good ones were priced out of a student project. I built it over a month in October 2025, about 15 hours of hands-on work, then retired it when the course ended. What it proved is the workflow, not the product: an AI tool can carry the first draft of a build, and the designer’s job becomes scoping, editing, and catching what the model gets wrong.