Cook or Cooked

Inspiration

As broke and slightly irresponsible university students, too often do we underestimate the amount of money we throw away on food - especially when eating outside. Every month ends with a budgeting crisis and many questions about where all the money went. Clearly, we need a smarter way to stay on top of our food expenses. Introducing Cook or Cooked, an app to track your spending on food with nothing but a picture of it. Visualize and let it sink in: how much money do you save (or waste) every time you choose to cook or eat outside?

What it does

Our app makes managing food spending as easy as humanly possible. Just take a picture of your meal and all the analysis will be done for you. With image recognition, price references from our database, and real time analytics, our app identifies what's on your plate and estimates the cost based on average market prices in your area. If you ate out: the app shows you how much it cost and how much you could’ve saved by cooking it yourself. If you cooked: it estimates the total cost of ingredients used and compares it to the price of a similar meal at a restaurant—letting you see your savings in real-time. You can track your weekly/monthly food expenses, spot trends (like how often you cave into ordering takeout), and set goals for smarter spending.

How we built it

Image Recognition: we integrated an image recognition model to identify food items from user-uploaded pictures. This serves as the starting point for both cost and nutrition analysis.
Food pricing database: we scraped online sources to build a dataset of food prices and nutritional information, enabling accurate comparisons between homemade and restaurant meals.
Supabase: we used Supabase as our backend database to store user data, meal logs, price estimates, and analysis results securely and efficiently.
Retrieval-Augmented Generation (RAG): to pull relevant data (e.g., average food prices, nutritional info) from custom sources, allowing the app to provide contextualized insights and comparisons.
LangChain + LangGraph: for building and orchestrating LLM Agents, helping us manage the flow of data between image recognition, retrieval, and output generation.
Cohere, Groq, and OpenAI: we leveraged multiple LLMs for different components - we used Cohere web search to assist the construction of our food pricing database, we used Groq for fast inference speeds for our components, and OpenAI for generating detailed insights and summaries in our specified output structure.

Challenges we ran into

The biggest challenge was landing on this idea. Many hours were spent going back and forth, trying to find something relevant to our lives. We knew we wanted to tackle a problem students actually face, but it took time to refine our direction. Once we landed on Cook or Cooked, that was when we were cooking.

Accomplishments that we're proud of

Coming up with something that's actually useful for us students—Cook or Cooked solves a real problem we face daily.
Creating an intuitive and appealing UI that makes tracking food spending simple and fun.

What we learned

We spend way too much money on takeout
How to integrate and optimize multiple APIs to reduce latency while maintaining output quality
How to handle and combine structured outputs from different components - merging data from image recognition, pricing databases, and nutrition sources into a coherent result.

What's next for Cook or Cooked

Nutrition Tracking: When it comes to food it's not just what it costs, but also what it's made of. By comparing nutrients across home-cooked versus restaurant meals, users can get insights into their consumption of sodium, sugar, preservatives, and other additives that are often overlooked when eating out.
Social Features + Accountability: Budgeting and eating healthy is much easier (and more fun!) when you do it with friends. Social features like friend comparisons, cook-at-home streaks, and group challenges make saving your wallet a fun and friendly competition.
Groceries Integration: Once you (finally) decide to cook at home, we want to make it as easy as possible. By integrating grocery platforms, users can automatically generate shopping lists based on their meal preferences and order directly through the app. This greatly reduces the friction between planning to cook and actually cooking.