AutoRescue

Architecture Diagram

Inspiration

As a frequent business traveler, I've experienced the nightmare of flight cancellations firsthand. Picture this: you're at JFK airport at 11 PM, your connecting flight to an important client meeting just got cancelled, and you're staring at a departure board full of "CANCELLED" notifications. The airline's customer service line has a 3-hour wait, their app crashes under load, and you're left frantically refreshing booking sites trying to find alternatives while your phone battery dies.

This scenario plays out millions of times each year. Flight disruptions affect over 20% of all flights, stranding travelers and costing airlines billions in rebooking fees and customer compensation. Yet the tools available to handle these crises are painfully outdated - static websites, overwhelmed call centers, and manual rebooking processes that take hours when minutes matter.

I realized there was a massive opportunity to revolutionize crisis management in travel using AI. While we have sophisticated AI assistants for everything from shopping to entertainment, travelers facing urgent disruptions are still left to navigate complex rebooking scenarios alone. The travel industry needed an intelligent agent that could instantly analyze disruptions, search alternatives, and execute rebookings with the speed and precision that crisis situations demand.

What it does

AutoRescue is an AI-powered flight disruption management assistant built on AWS Bedrock AgentCore that transforms travel crisis response from a manual nightmare into an automated, intelligent process. It provides:

Intelligent Flight Search: Real-time flight discovery across multiple dates and carriers using live Amadeus API data, with advanced filtering for disruption scenarios.

Smart Disruption Analysis: Analyzes cancelled flights and automatically generates comprehensive rebooking strategies, considering factors like price, convenience, timing, and user preferences.

Real-time Pricing: Validates flight offers and provides accurate, up-to-the-minute pricing including all taxes and fees through direct Amadeus integration.

Autonomous Crisis Management: Handles the entire rebooking workflow from disruption notification to alternative recommendations with minimal human intervention.

Contextual Intelligence: Understands urgency levels, travel constraints, and user preferences to prioritize the most relevant solutions during high-stress situations.

The platform transforms chaotic travel disruptions into manageable, automated processes that save travelers hours of stress and help airlines reduce customer service load during crisis events.

How we built it

Technical Architecture:

Agent Framework: Built using AWS Bedrock AgentCore with native Claude Sonnet 4.5 integration
AI Models: Powered by Claude Sonnet 4.5 for advanced reasoning and decision-making in crisis scenarios
External Integration: Production-grade Amadeus Flight API integration with OAuth2 authentication
Serverless Backend: AWS Lambda functions for scalable, independent tool execution
Security: AWS Secrets Manager for credential storage, Cognito for authentication

Data Pipeline:

Flight Search: Real-time queries to Amadeus Flight Offers API with carrier filtering and multi-day search capabilities
Pricing Validation: Complete flight offer pricing through Amadeus Pricing API with tax and fee calculations
Disruption Analysis: Intelligent analysis of cancelled flights with alternative recommendation algorithms
Session Management: AgentCore Memory for conversation persistence and context retention

Lambda Architecture:

SearchFlights: Comprehensive flight search with advanced filtering (200+ lines of production code)
OfferPrice: Real-time pricing validation with complete Amadeus object handling
BookFlight: Flight booking framework (ready for production deployment)

User Experience:

Streamlit Interface: Modern, responsive web application with authentication
Crisis Simulation: Realistic flight cancellation scenarios with locked parameters
Real-time Updates: Live flight data with immediate pricing and availability

Challenges we ran into

API Complexity: Amadeus Flight API has intricate data structures and strict validation requirements. Understanding the complete flight offer objects needed for pricing took significant debugging, especially handling mandatory fields like travelerPricings and segment IDs.

Authentication Architecture: Implementing secure OAuth2 flows for both Amadeus API and AgentCore Gateway while maintaining credential caching and token refresh logic across multiple Lambda functions.

AgentCore Integration: Learning AWS Bedrock AgentCore's native SDK and deployment patterns, moving beyond wrapper frameworks to direct integration with the AgentCore runtime.

Real-time Performance: Ensuring sub-5-second response times for flight searches while handling API rate limits, network timeouts, and credential refresh cycles.

AI Reasoning: Training Claude to understand the urgency and constraints of flight disruption scenarios, balancing speed with thoroughness in recommendation generation.

Data Validation: Handling inconsistent flight data formats, missing terminal information, and edge cases in pricing calculations while maintaining user-friendly error messages.

Accomplishments that we're proud of

Production-Grade Integration: Built a complete, working flight booking workflow with real Amadeus API integration - not just mock data or simulations.

Native AWS Architecture: Achieved true AWS-native implementation using Bedrock AgentCore SDK directly, demonstrating deep integration with AWS services.

Crisis Management Innovation: Created the first AI agent specifically designed for flight disruption scenarios, with intelligent reasoning about urgency, constraints, and user priorities.

Sophisticated Lambda Functions: Developed production-quality serverless functions with comprehensive error handling, logging, and performance optimization.

Enterprise Security: Implemented proper credential management, OAuth2 flows, and secure API integration patterns suitable for production deployment.

Realistic User Experience: Built an authentic crisis simulation that demonstrates real-world value, complete with locked parameters and time pressure scenarios.

What we learned

AgentCore Mastery: Gained deep expertise in AWS Bedrock AgentCore's native capabilities, understanding the difference between wrapper frameworks and true AWS integration.

Travel Industry Complexity: Discovered the intricate world of airline APIs, booking workflows, and the technical challenges behind seemingly simple flight searches.

AI Crisis Management: Learned how to design AI reasoning for high-stress, time-sensitive scenarios where traditional chatbot approaches fall short.

Serverless Orchestration: Mastered the art of coordinating multiple Lambda functions through AgentCore Gateway while maintaining performance and reliability.

Real-time Data Challenges: Understood the complexities of working with live, constantly-changing flight data and the importance of caching strategies.