Inspiration
Most people don’t eat badly because they don’t care. They eat badly because they don’t know what their meals actually mean for their health. Is my lunch balanced? Did I eat too much sodium this week? Is this meal actually helping my goal? Nutrition apps usually just count calories. I wanted to build something that actually explains your food in plain language and helps you improve it without making you feel judged. That’s how AlternAte started.
What it does
AlternAte is an AI-powered nutrition assistant. You can type any meal, like: “Grilled chicken sandwich with fries and iced tea.” In seconds, AlternAte shows: Calories Protein, carbs, fat Fiber, sugar, sodium Health flags (like high sodium or low fiber) It also gives you 3 smarter swaps based on your goal, whether that’s weight loss, muscle gain, heart health, or just eating balanced.
Compare Two Meals
You can compare two meals side by side. For example: A burrito bowl vs. a loaded quesadilla. AlternAte breaks down both meals and explains: Which one better fits your goal The tradeoffs How to improve either one It doesn’t just show numbers. It tells you what actually matters.
Check Your Whole Diet
You can paste your entire day of eating, and AlternAte will: Score it from 0–100 Tell you what’s working Point out what’s too high or too low Suggest a realistic improved version of your day You can also upload a food photo for analysis.
Meal Log and Weekly Score
Every meal you analyze can be saved. Your saved meals build a weekly health score based on things like: Sodium control Sugar control Fiber intake Protein balance Calorie balance Variety The score updates automatically and gives you a tier rating from low to very good.
Leaderboard and Points
AlternAte also makes healthy eating social. You can compete with friends on a weekly leaderboard. The better your score, the higher you rank. You earn points for actions like: Saving meals Completing diet checks Logging consistently Reaching high weekly scores It turns nutrition into something interactive instead of overwhelming.
How we built it
The app is built with React and Vite. All AI calls go through a secure Vercel serverless function so the API key is never exposed to users. I use two Gemini models: One for full analysis, comparisons, and vision One lightweight version for fast meal breakdowns All responses are forced into strict JSON so the data stays structured and reliable. Meal history and points are stored locally in the browser for the prototype.
Challenges we ran into
Making AI responses consistent and parseable Preventing long diet-check responses from getting cut off Designing a scoring system that feels fair and realistic
Accomplishments that we're proud of
Prompt design matters more than expected Serverless deployment is powerful and secure Gamification genuinely changes how people think about their habits
What we learned
Most people don’t want more data, they want clear explanations. Small, realistic swaps are more powerful than extreme diet rules. A scoring system makes progress visible and motivating. AI works best when prompts are tightly structured and outputs are controlled. Social features make healthy habits feel less isolating and more engaging.
What's next for AlternAte
Improve long-term health projections with stronger nutrition modeling Expand food recognition and improve photo analysis accuracy Add personalized trend reports based on longer meal history Introduce streak-based challenges and goal tracking Launch a mobile-optimized version Test with real users and refine the scoring system based on feedback
Log in or sign up for Devpost to join the conversation.