Foodora

Foodora is an AI-powered app that tells restaurant owners exactly how much food to prepare each day. It predicts demand, prevents overproduction, and routes surplus to food banks.

Comment

Project Story

The Problem

Food waste is a major issue in the restaurant industry. Restaurants must prepare food in advance without knowing exactly how many customers will arrive. Because of this uncertainty, many restaurants intentionally overprepare to avoid running out of items. While this protects sales, it often results in significant food waste at the end of the day.

We realized that the core problem is demand uncertainty. If restaurants could better predict how much food they will sell on a given day, they could prepare closer to the right amount and dramatically reduce waste.

This insight inspired us to build a system that helps restaurants make data-driven preparation decisions.

⸻

Our Solution

We built an AI-powered restaurant assistant that predicts daily demand for menu items and helps restaurants reduce overproduction.

Restaurant owners can upload historical sales data or manually log sales throughout the day. Our system combines this information with contextual factors such as weather conditions, location type (e.g., mall, university, or downtown area), and upcoming holidays to estimate how many units of each menu item will likely sell.

For example, a restaurant might see a prediction like:

\text{Predicted Shawarmas Today} = 95

Using this information, the restaurant can prepare food in more accurate quantities instead of relying on guesswork.

Throughout the day, sales are logged and the system continuously updates its understanding of demand. At closing time, the restaurant can close the store in the dashboard and the system calculates overproduction and waste. If excess food remains, the app suggests nearby donation centers so the food can still serve the community.

⸻

How We Built It

The application is designed as a mobile-first platform focused on restaurant operators rather than consumers.

Our system architecture includes: • Mobile App Interface for restaurant dashboards and daily insights • AI Prediction Layer powered by Google Gemini • Backboard.ai Agent Framework to manage AI workflows and extended memory • Voice Interaction Layer using ElevenLabs to enable voice-based sales logging • Structured Data Inputs such as historical sales logs and CSV uploads

Backboard plays a key role by enabling an AI agent that can maintain contextual memory about restaurant operations. Instead of generating isolated predictions, the system can analyze historical behavior and provide more informed recommendations over time.

The voice interface allows workers to log sales naturally by speaking orders such as:

“Three chicken shawarmas and two falafels.”

The AI interprets the order and updates the system automatically, making sales tracking significantly faster and more convenient in a busy kitchen environment.

⸻

Challenges We Faced

One of the biggest challenges was determining how to structure restaurant data so the AI could generate useful predictions. Restaurants generate many different types of signals (sales logs, location patterns, weather effects), and organizing these inputs in a meaningful way required careful design.

Another challenge was making the system practical for real restaurant workflows. Initially, manual logging of sales felt inefficient. To solve this, we introduced a voice-based logging system that allows staff to record orders without interrupting their workflow.

We also explored how to use AI agents effectively. Instead of simply generating predictions, we wanted the AI to provide operational insights such as identifying waste patterns or explaining why sales were higher or lower on certain days.

⸻

What We Learned

Through this project we learned how AI can be used not just for analysis but for operational decision-making. Restaurants do not simply need predictions—they need tools that translate predictions into actions.

We also learned the importance of designing AI systems that integrate naturally into real-world workflows. Features like voice logging and automated insights help bridge the gap between complex AI systems and everyday restaurant operations.

Most importantly, this project showed us how emerging AI tools and agent frameworks can be used to solve practical sustainability problems. By helping restaurants prepare food more accurately and redirect excess food to community organizations, our platform aims to reduce waste while improving operational efficiency. What’s Next for Fodora

In the next 30 days, our goal is to make Fodora even easier for restaurants to adopt and integrate into their daily operations.

One of our top priorities is building automatic POS integrations so restaurants no longer need to manually log sales. By connecting directly with existing point-of-sale systems, Fodora will automatically receive real-time sales data, allowing the AI to continuously update predictions and provide more accurate preparation recommendations.

We also plan to expand the AI assistant’s capabilities by adding deeper operational insights. This includes features like identifying long-term waste patterns, recommending optimal batch preparation sizes, and providing smarter daily summaries that help restaurant owners understand why demand changed on a given day.

Another key improvement will be refining the voice-based logging system, allowing kitchen staff to quickly record sales or inventory updates through natural conversation, making the platform even more convenient in busy restaurant environments.

Our long-term vision is to evolve Fodora into an AI copilot for restaurant kitchens, helping businesses reduce food waste, improve operational efficiency, and make smarter decisions using data-driven insights.

Built With

Share this project:

Updates