Airline Disruption Supplychain Manager

Starting scenario
see all flight scheduled for next 12hrs
Select airline
All flights sudden failure Cancellation dummy scenario execution

Inspiration

Flight disruptions create chaos for both passengers and airline teams. During large-scale cancellations, decisions are manual, slow, and stressful. We wanted to build something that helps airlines respond faster and smarter when things go wrong.

What it does

Airline Disruption Supplychain Manager monitors flight disruptions in real time, groups passengers by urgency, and automatically finds the best rebooking options using live flight data. It helps airlines take clear, prioritized recovery actions instead of reacting late.

How we built it

We used Confluent Kafka to stream real-time flight, booking, and passenger data. SQL stream processing calculates disruption pressure and passenger urgency. A backend service integrates with Amadeus to fetch alternate flight options, and everything is hosted as a web app with a live dashboard.

Challenges we ran into

Setting up real-time stream processing correctly was challenging, especially choosing the right infrastructure. Handling external API limits and keeping the system responsive under disruption scenarios also required careful design.

Accomplishments that we're proud of

We built a working end-to-end real-time system that reacts to disruptions instantly. Passenger prioritization, live rebooking options, and a clean dashboard all work together in a realistic airline recovery flow.

What we learned

We learned how powerful event-driven systems are for operational problems. Real-time data, when combined with clear business rules and APIs, can drastically improve decision-making during critical situations.

What's next for Airline Disruption Supplychain Manager

Next, we plan to add hotel and ground transport handling, deeper AI-driven decision-making, and integrations with airline operations systems to move closer to real-world deployment.

Built With

amadius
confluent
fastapi
gemini-flash-3.5
mapbox
python
react

Submitted to

AI Partner Catalyst: Accelerate Innovation

Created by

I led the backend architecture and AI agent design for the Airline Disruption Supplychain Manager, focusing on building an enterprise-style, production-ready decision system rather than a simple demo.

My key contributions include:

Designed and implemented a multi-agent orchestration pipeline to handle airline disruption recovery end-to-end. This includes:

A Triage Agent to assess disruption severity and derive operational constraints.

A Rebooking Agent to normalize, filter, and validate alternative flight options from live airline data.

A Decision Agent powered by an LLM (Claude) to rank recovery options, generate human-readable reasoning, and assign confidence scores—mirroring how real operations teams justify decisions.

Integrated real airline data using the Amadeus API, building a robust normalization layer to convert raw flight-offer responses into internal schemas suitable for AI reasoning and ranking.

Built a FastAPI-based backend service with clean API boundaries, orchestration logic, and error handling, enabling real-time recovery recommendations via a single endpoint.

Implemented hybrid AI design principles, deliberately combining:

Deterministic, rule-based logic for safety-critical steps (triage, constraint enforcement)

LLM-driven reasoning only where judgment, explanation, and trade-off analysis are required
This mirrors how real-world airline and logistics AI systems balance reliability, cost, and explainability.

Handled infrastructure and system reliability, including environment configuration, secure API key management, service health checks, and local simulation endpoints for testing disruption scenarios.

Overall, my work focused on turning airline disruption management into an explainable, agent-driven decision system, rather than just surfacing flight options—laying the foundation for scalable, event-driven recovery workflows used in real airline operations.

Mahima Advilkar
Ashu Kumawat

Updates

Ashu Kumawat started this project — Dec 31, 2025 04:59 PM EST

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.