My Contribution
As the sole developer and project architect, I was responsible for designing, building, integrating, and deploying the entire WanderWise AI system — a fully functional, Agentic AI-powered travel assistant built end-to-end on Google Cloud.
1. Concept and Architecture Design
I conceptualized WanderWise AI as an Agentic AI travel planner that plans like a human — reasoning, adapting, and learning from user input.
I designed the complete system architecture connecting Gemini Pro, the Agent Development Kit (ADK), Firebase AI Studio, Cloud Run, and Firestore into a unified, serverless ecosystem.
I also created detailed architecture and data flow diagrams to visualize how the AI, backend, and frontend interact in real time.
2. Backend Development (FastAPI, ADK, Cloud Run)
I developed the backend from scratch using Python 3.13 and FastAPI.
This included routes for itinerary generation, authentication, and data storage.
I integrated Google’s Agent Development Kit (ADK) to create a custom itinerary_agent capable of generating structured, day-by-day plans based on user preferences.
I containerized and deployed the backend on Google Cloud Run using Docker and Cloud Build for CI/CD automation.
The deployed backend achieved zero percent error rate, one hundred percent uptime, and a consistent response time under ninety milliseconds.
3. AI Integration and Prompt Engineering
I connected Gemini Pro to the backend using the Genkit SDK and ADK Runner.
I designed structured prompt templates to generate reliable JSON responses, ensuring accuracy and consistency across outputs.
I implemented context management logic so that Gemini could maintain continuity and reasoning across multi-turn travel conversations.
4. Frontend Development (Firebase AI Studio and Next.js)
I built the entire frontend inside Firebase AI Studio using Next.js (version 15) and TypeScript.
The interface includes a conversational chat layout that interacts with the backend APIs for real-time itinerary generation.
I added multilingual support for English, Hindi, and Kannada, and styled the user interface using Tailwind CSS and ShadCN UI.
The frontend was deployed through Firebase App Hosting for a fast, secure, and serverless experience.
5. Data and Authentication Layer
I configured Firebase Authentication to enable secure Google and Email/Password login options.
I connected Firestore for persistent data storage to maintain user itineraries, preferences, and session history.
I also implemented a demo mode using environment variables to ensure safe and consistent demonstrations.
6. Design, Branding, and User Experience
I created the WanderWise AI brand identity, using a muted green and warm beige color palette for a calm, mindful travel theme.
The interface follows a “watercolor and storybook” aesthetic for an inviting and professional appearance.
I designed all user flows and ensured that the experience is accessible across both desktop and mobile devices.
7. Testing, Deployment, and Documentation
I performed full system testing across all components, including backend health checks, API latency verification, authentication flow, and real-time data connections.
I monitored performance through Cloud Run Observability and Cloud Logs Explorer.
I prepared detailed documentation, including the project report, architecture diagrams, and setup guide for reproducibility.
8. Overall Impact
I built a production-ready AI system entirely using Google Cloud’s native technologies.
The project demonstrates how Agentic AI can be applied to real-world travel planning through reasoning, adaptation, and multimodal interaction.
It also provides a scalable foundation for future expansion into booking systems, analytics, and multimodal interfaces.
Summary of My Role
I led the entire project from ideation to deployment, including system architecture, backend and frontend development, AI integration with Gemini and ADK, serverless deployment on Cloud Run, and user experience design.
WanderWise AI reflects a complete implementation of an Agentic AI ecosystem that is practical, scalable, and production-ready.
Log in or sign up for Devpost to join the conversation.