Nexus Gate Predict

Gate Group wastes millions losing 50% of inventory. Our ML and computer vision system cuts their 4-hour process to under 1 hour. 60K flights/day. The future of airline logistics is here today.

Comment

About the Project: Onboard Intelligent Inventory Assistant

Inspiration

This project was born from identifying a common problem in airline operations: crew members often struggle to locate products, check expiration dates, and record consumption during flights — especially without internet access. The idea was to create an intelligent assistant that works completely offline, providing instant access to all inventory information while anticipating needs and suggesting actions.

What We Learned

Throughout the development process, we deepened our understanding of:

  • Offline-first architectures with PostgreSQL
  • Data synchronization between the cloud (Snowflake) and mobile devices
  • Conversational interfaces for low-connectivity environments
  • Predictive models for consumption and expiration
  • User-centered design for high-pressure contexts

We also explored computer vision for product recognition and how to structure a local database optimized for fast queries.

How We Built It

The system is composed of three main phases:

1. Pre-Flight Synchronization (Online Mode)

  • Snowflake generates the complete flight plan:

    • Passengers, trolleys, products, expiration dates

  • All information is downloaded via REST API

  • It is stored in a local PostgreSQL database on the tablet or smartphone

2. In-Flight Operation (Offline Mode)

  • All queries are resolved locally
  • The assistant answers questions like “Where’s the Coca-Cola?” or “What should I use first?”
  • Inventory updates in real time with each consumption
  • Alerts are generated for products nearing expiration
  • Products can be identified via QR scanning or computer vision

3. Post-Flight Synchronization

  • Upon detecting Wi-Fi after landing:

    • Consumption logs, alerts, waste, and metrics are uploaded
    • Data becomes available for analysis and to improve future operations

Technologies Used

  • Backend: Snowflake, REST APIs
  • Frontend: Mobile application (Flutter)
  • Local database: PostgreSQL
  • AI/ML: Offline vision models, predictive analytics
  • Conversational interface: Assistant with natural language processing

Challenges Faced

  • Designing a fully functional offline system
  • Maintaining data consistency between devices and the cloud
  • Handling exceptions such as damaged products or last-minute changes
  • Optimizing performance on resource-limited devices
  • Creating a natural and efficient conversational experience

What Makes It Special

  • Works 100% offline during the flight
  • Real-time inventory tracking
  • Intelligent recommendations based on expiration and consumption
  • Visual product recognition
  • Predictive analytics to enhance future flight planning

This project is more than just a tool — it’s a digital copilot for the crew, designed to make their work easier, faster, and more accurate.

Share this project:

Updates