UniBite

Inspiration

Our project was inspired by the “Freshman 15,” a common term referring to the weight gain some students experience during their first year of college. For many students, college is the first time they have complete control over what they eat, and dining halls offer a huge number of choices every day. We wanted to make those choices easier by helping students build balanced meals based on their goals and preferences. Rather than focusing only on short-term weight management, our app encourages healthy eating habits that can support long-term health and longevity.

What it does

UniBite turns the question “What should I eat today?” into a personalized meal plan using food that’s actually available on campus.

Students simply describe what they want in plain English—like “Plan me a high-protein day under 2,000 calories” or “Give me healthy meals for the week.” UniBite pulls live menu and nutrition data from the university's dining halls and builds a meal plan around the student’s calorie and macro goals, dietary preferences, and available options.

Users can view their plan by day, see nutritional totals, and refine it conversationally if something doesn’t fit, whether they want more protein, fewer calories, or just a different meal. Plans are saved with revision history, so students can easily revisit and adjust them.

How we built it

We built UniBite as a full-stack web application that connects directly to the University of Houston’s DineOnCampus menu data. Our FastAPI backend fetches available dining hall items and their nutritional information, then evaluates different meal combinations based on a user's calorie and macro goals.

On the frontend, we used React and TypeScript to create a simple weekly meal-planning experience. Instead of making users manually enter every requirement, UniBite accepts natural-language requests such as “Plan my meals for tomorrow with high protein and around 2,000 calories.” The request is translated into nutritional targets and constraints, and UniBite generates meals using food that is actually available on campus. We also integrated Gemini as an optional layer for understanding and refining natural-language requests.

Challenges we ran into

One of our biggest challenges was scalability. UniBite currently uses live dining data from the University of Houston, but our goal is to make it work for students at any university. Different universities use different dining platforms and expose their menu and nutrition data in different formats—if they make that data accessible at all. We had to think and design UniBite in a way that could eventually support multiple campus dining systems and data sources.

Another major challenge was frontend design and usability. Nutrition and meal planning involve a lot of information such as calories, dietary restrictions, dining locations, individual food items, daily and weekly plans. We had to find a balance between showing users everything they need while keeping the experience simple, intuitive, and easy to understand that they could realistically use it every day.

Accomplishments that we're proud of

We're proud that UniBite goes beyond being another AI meal recommender. The meals it suggests are grounded in real food that students can actually find at dining halls. We also created an experience where students can describe what they want conversationally and then continue refining it without rebuilding their entire plan. Most importantly, we connected an everyday student decision to the bigger idea of longevity. UniBite encourages students to start building healthier eating habits early rather than waiting until health becomes a problem.

What we learned

The experience taught us that creating a useful health application requires more than integrating an AI model. We learned how to combine live external data, nutritional calculations, natural-language processing, and constraint-based decision making into one system.

We also learned the importance of putting guardrails around generative AI, especially when dealing with health-related recommendations. Gemini can help understand what the user wants, but our application logic remains responsible for validating nutritional values and ensuring recommendations come from actual dining hall options.

What's next for UniBite

Our biggest next step is scaling UniBite beyond the University of Houston. We want to integrate dining menu and nutrition data from universities across the country, allowing students to get personalized meal plans wherever they are on campus.

We also plan to expand personalization with saved dietary preferences, allergies, fitness goals, and more detailed nutrition factors beyond macros. Over time, UniBite could track eating patterns and help students understand how their everyday food choices contribute to their long-term health and healthspan.