Inspiration

Our inspiration started at our dinner tables. We wanted to get serious about our fitness, and we noticed that many of the things our families were cooking didn’t align with our fitness goals. The content creators who inspired us had one answer, which was to start fresh and eat just chicken, rice, and broccoli. That nuclear option was never realistic for us, as we didn’t run our kitchens, and we were not willing to give up our families' foods.

What it does

Famara recommends better-for-you versions of the food you already love, and tells you why each one fits you.

  • Discover. A swipe deck drawn from 1,031 recipes across 60 cuisines, ranked for your cuisines, goals, time and allergens. Every card shows a fit score and the top two reasons it fits. It learns from what you like, skip and cook.
  • Make it healthier. Pick a goal such as more protein or lower sodium and Famara adapts the recipe. Compare shows the original beside the new version, ingredient by ingredient and nutrient by nutrient.
  • Snap. Photograph your pantry, or film one slow pass of the fridge and say what the camera missed. Meals for now then shows what you can cook right away, sorted by time of day. Snap a finished dish to get estimated nutrition and a healthier version.
  • My recipes. Bring in your family's recipes by link, a photo of a recipe card, by voice, or by typing. Famara checks every ingredient before it suggests the dish back to you.
  • Cooking coach. Rest your phone on a cabinet shelf above the counter and cook hands-free. The screen reads from across the room. Say "Hey Famara, set a 2 minute timer for the peas", or ask "is this stirred enough?" and it takes one photo and answers out loud. No video is streamed or stored.
  • Plan and shop. Auto-fill a week of meals around your goals, time and pantry. The shopping list merges ingredients, subtracts what you own, groups by aisle and estimates cost.
  • Chef, Community and Nearby. An AI chat that knows your profile, a feed of real people's cooked dishes and remixes, and a map of restaurants likely to serve the dish you are craving.

Safety is built in, not bolted on. Allergen filtering is fail-closed: if Famara cannot confirm a recipe is safe for an allergen you declared, you never see it. Anything the AI recommends is re-checked on your device before it is shown. Nutrition is labelled as computed from USDA data or estimated. AI features are limited to adults who have given consent.

How we built it

-React Native 0.86 with Expo SDK 57 and strict TypeScript, so one codebase targets iOS and Android. We polished iOS first, with a Liquid Glass interface in light and dark mode. -Our own data pipeline builds the recipe catalog from TheMealDB, MedlinePlus and USDA FoodData Central, then ships it inside the app as a read-only SQLite database. Search and recommendations run on the device, so they are instant. -Backend is through Firebase -381 test files, database tests in pgTAP, and CI running typecheck, lint, format and tests on every pull request. -We spent the first half of the week planning every screen, feature and user flow. The first commit landed on day four. From there we orchestrated AI coding agents (Claude Code and OpenAI Codex), running subagents in parallel on separate workstreams and using a second model family to check the first one's work. That took us through about 70 pull requests in five days.

Challenges we ran into

One challenge that we ran into was E2E testing; due to the limited time, we wanted our agents working around the clock, but they kept finishing before reaching our expected quality level. So we ended up dedicating a large chunk of time to explaining to the agents what our quality meant, and worked one-on-one to fix all the minor UI issues, the ones impossible to miss once noticed. Building our safety system was also a challenge that we spent a considerable amount of time on, as a model suggesting a dish you are allergic to can be extremely dangerous. To solve this, we made the catalog the source of truth, and rechecked every AI suggestion locally before the user sees it.

Accomplishments that we're proud of

We are quite proud that we were able to ship a production-ready app, not just a demo of one screen but a complete suite of discover, adapt, plan, shop, cook, and share features. A cooking coach you can talk to with your hands covered in flour. An allergen system that says "we don't know" instead of guessing. The ability to bring recipes that were passed down for generations or one that you really like eating into the app, and refined for your health and fitness goals while retaining taste. Finally, the skills we take away from building something this large in one week.

What we learned

Over the course of this hackathon, we learned the plan-first approach. We went against our instincts to immediately begin vibecoding and instead fully thought out every screen, feature, and user flow before the first line of code was written. With the amount of interconnected features, the project likely wouldn’t finish otherwise. We will certainly take this approach in all of our future endeavors. Alongside this, we became much better orchestrators. We used subagents, workflows, and different model families working in tandem to complete our project on time.

What's next for Famara

We want to bring Famara to real families, and it will begin with our own. This means submitting the app to the App Store and the Google Play Store.

Built With

Share this project:

Updates

Submission history