AI Travel Recommendation App

GitHub:

Value statement: Serverless full-stack application delivering personalized travel recommendations using AWS Bedrock and Claude 3 Sonnet with GraphQL API and Cognito authentication.

Overview

Built a serverless travel recommendation app leveraging AWS Bedrock’s Claude 3 Sonnet for intelligent destination suggestions. Users input interests and receive personalized travel recommendations with destinations, activities, and itineraries.

The application demonstrates modern cloud-native architecture: React/TypeScript frontend hosted on AWS Amplify, AWS AppSync GraphQL API, Lambda functions for AI inference, and Amazon Cognito for authentication. Fully serverless design with Git-based CI/CD workflow.

Architecture

AI Travel App Architecture Additional Image

Technology Stack

Layer	Technologies
Frontend	React 18, TypeScript, Vite, Amplify UI
Authentication	Amazon Cognito (Amplify Auth)
API	AWS AppSync (GraphQL)
Backend	AWS Lambda
AI/ML	AWS Bedrock (Claude 3 Sonnet)
Hosting	AWS Amplify (CloudFront + S3)
CI/CD	AWS Amplify Git-based workflow

Implementation Details

AWS Bedrock + Claude 3 Sonnet: Lambda function invokes Bedrock with prompt engineering for travel recommendations:

response = bedrock_runtime.invoke_model(
    modelId='anthropic.claude-3-sonnet-20240229-v1:0',
    body=json.dumps({
        "anthropic_version": "bedrock-2023-05-31",
        "messages": [{"role": "user", "content": prompt}],
        "max_tokens": 2000
    })
)

AWS AppSync GraphQL API:

Custom queries reference Lambda data source
Resolvers process data between frontend and Bedrock
Structured response types for AI model outputs

React/TypeScript Frontend:

Vite build tooling for fast development
Amplify UI components for authentication flows
TypeScript for type-safe API contracts
Responsive design for mobile/desktop

Amazon Cognito (Amplify Auth):

User registration with email verification
Sign-in/sign-out with JWT tokens
Password reset workflows
Automatic token refresh

Amplify CI/CD: Git-based workflow with automatic deployments on push to main, preview environments for PRs, and global CDN distribution.

Features

Personalized recommendations based on user interests
AI-generated itineraries with destinations and activities
Secure authentication with sign-up, sign-in, password reset
Responsive UI with Amplify component library
Serverless scaling handles variable traffic automatically

Performance & Cost

Metric	Value
Response Time	2-4 seconds (Bedrock inference)
Concurrent Users	1000+ (Lambda auto-scaling)
Monthly Cost	~$15-30 (low usage)
Bedrock Cost	$0.015 per 1000 input tokens

Lessons Learned

Prompt engineering is critical: Initial prompts produced verbose, unstructured responses. Iterative refinement with output format specifications improved quality significantly.

Claude 3 Sonnet vs Haiku: Started with Haiku for cost savings but switched to Sonnet for better reasoning. Cost increase (2x) justified by quality improvement.

Amplify simplifies DevOps: Compared to manual CloudFront/S3/Lambda setup, Amplify reduced deployment complexity by 80% while providing preview environments and atomic deployments.

Future Improvements

Implement RAG with vector database for real-time travel data
Integrate travel APIs for pricing and availability
Add user history for personalized recommendations
Support multiple languages with Claude’s multilingual capabilities