Decision system inputs
Getting AI Description
AI Generated Description
Simulation pick & pack
Results for two different types of cookie packages

Picky Packers

Inspiration

GateGroup handles hundreds of products daily for different airlines. When errors happen, nobody knows where or why. We built a system that breaks down complex workflows into digital, measurable steps.

What it does

Module 1: Alcohol Compliance

Input flight number and bottle conditions
Gemini AI analyzes against airline policy
Get instant decision: KEEP, REFILL, REPLACE, or DISCARD with reasoning

Module 2: Production Line Assistant

Shows step-by-step instructions
Validates with weight sensors (simulated)
Computer vision model recognizes items
Alerts immediately on errors
Auto-advances when correct

How we built it

React + TypeScript frontend
Google Gemini 1.5 Pro for AI decisions
Computer vision model for object recognition
Google Cloud Platform + OAuth
Aviation Edge API for flight data
Simulation environment for workflow testing

Challenges we ran into

Training computer vision to recognize items accurately
Designing weight validation logic that handles real-world scenarios
Creating UI that's fast enough for production workers

Accomplishments that we're proud of

AI compliance decisions in 10 seconds
Working computer vision model that identifies items
Weight-based validation simulation
Clean modular architecture ready for integration

What we learned

Building modular systems from the start makes everything easier. Real-time feedback systems need to be fast and clear. Combining multiple validation methods (weight + vision) creates robust error detection.