GateSort

Inspiration

In the high-pressure environment of airline logistics, time is money and small mistakes can have major consequences. Expired food or mismanaged alcohol can lead to safety risks and regulatory issues. Crew members often perform these checks manually, under time pressure and fatigue. Aditionally rules change according to the regulations of each airline.

Therefore we wanted to build a solution that takes the mental load off staff while increasing accuracy, speed, and compliance.

What it does

Gate Sort is a digital solution to enhance the efficiency of airline catering logistics. Designed for airline crew and support staff, the platform focuses on minimizing time and cognitive load when checking onboard inventory, particularly food expiration and alcohol bottle levels. The system integrates AI and computer vision to automate two critical tasks: Food Expiry Verification

Using a system of printed stickers and visual AI, Gate Sort scans food items and flags them as Safe, Near Expiry, or Expired. This allows crews to quickly identify items that should not be loaded onto flights, ensuring passenger safety and regulatory compliance.

Users capture multiple images of alcohol bottles, which are analyzed by AI to determine whether the contents are Full, Over Half, or Under Half. Based on the selected airline's policy, Gate Sort provides clear, actionable guidance on whether to reuse, discard, or flag the bottles for manual inspection. Alcohol Bottle Level Detection

How we built it

Design Principles: Built with a fatigue-friendly, minimalist interface tailored for tired crew members. Responsive, touch-friendly UI optimized for mobile and tablet use. Dark color scheme inspired by Gate Group’s brand identity, replacing gold with cool tones for accessibility. Technology Stack: React frontend using node js Integration-ready with AI/computer vision APIs for sticker and bottle analysis.

Challenges we ran into

How to read the expiration date one by one. Model training difficulties: The AI had trouble recognizing partial or damaged stickers, and sometimes misclassified them when similar shapes overlapped. Learning curve: This was our first time implementing computer vision, so understanding how to train and apply it took significant effort.

Accomplishments that we're proud of

Accurate AI Detection Our AI model can now reliably detect liquid levels in alcohol bottles, even under imperfect conditions such as glare, partial obstruction, or varied backgrounds.

Custom Sticker System for Expiry Tracking We developed a scalable and intuitive sticker labeling system. As told by the products arrive at the warehouse, and all items from the same lot share the same expiration date. The plan is to implement an automated or semi-automated labeling process where each product receives a sticker. The sticker’s color indicates the expiration month, and the icon or shape represents the expiration year, both of which are used by the AI model for visual recognition. Additionally, we aim to include the exact expiration day as a printed number on the sticker, allowing for greater precision and enabling just-in-time inventory control. This allows crew to quickly assess product freshness with computer vision. We plan to enhance this system by incorporating the exact expiration day, enabling just-in-time inventory control reducing waste, optimizing rotation, and improving catering efficiency. We successfully translated a complex manual process into a smooth digital workflow.

What we learned

Fundamentals of training computer vision models End-to-end flow of integrating AI into a React-based product Balancing usability and technical functionality for real-world fatigue scenarios

What's next for GateSort

Real-Time Video Analysis Enable live AI interaction through continuous video scanning. As the user shows each bottle, the AI will provide instant feedback on fill level and recommended actions. We would like to reduce friction and speeding up the process even further.

Sticker System Redesign Develop a cohesive, elegant, and standardized visual system for expiration stickers. This includes refining the iconography and color coding, ensuring it's universally readable and scalable across products.

Model Retraining for Granular Accuracy Improve the AI’s precision by training the model to detect not just the month and year, but the exact expiration date (day/month/year). This will allow for more accurate decisions, especially for items near their expiration threshold.