MedChef

Inspiration

Food can affect how some medicines work, but the warnings can be hard to connect to an everyday meal. We wanted to make that information easier to approach, especially for older adults, people managing long-term conditions, and their caregivers.

What it does

MedChef compares medication and food names against a small set of example interaction rules. Users can type their lists, speak their groceries in supported browsers, or read text from a photo of a label or menu.

It explains possible concerns in plain language and shows questions worth discussing with a pharmacist. For example, entering warfarin and spinach brings up why keeping vitamin K intake consistent matters, rather than telling someone to stop eating greens.

This is an educational prototype. An unflagged combination is marked “not assessed,” and the app does not tell users to change their medicine or diet.

How we built it

We built the interface with HTML, CSS, and vanilla JavaScript. The checker matches food names and generic or brand-name medicines against 13 example rules, using case-insensitive matching with word boundaries.

Tesseract.js reads text from photos in a browser worker, so the photos are not uploaded. Users review and correct the text before adding it to either list. The Web Speech API provides voice input where the browser supports it. The app also handles unreadable photos, denied microphone access, and missing inputs.

Challenges we ran into

A key challenge was explaining possible interactions without giving false reassurance or blanket food restrictions. “No match” needed to mean that the prototype had not assessed the combination.

Photo text recognition can also misread medicine names. We added a review step so users can correct the text before running a check.

Accomplishments that we're proud of

We brought typing, voice input, and photo text recognition into one simple flow. The results use everyday language, with FDA and NIAAA links for several examples. We also made the limits visible in the results themselves, including when no rule matches.

What we learned

Health information needs context. A food warning can depend on the exact medicine, eating habits, and a person's health. Clear wording, a chance to correct input, and honest limits are just as important as finding a match.

What's next for MedChef

Expand the interaction data with pharmacist review, test the interface with older adults and caregivers, and add a shareable summary for appointments. We also want to replace the fixed demo meal idea with meal planning that accounts for medicines, allergies, and individual care needs.