Voyager

The Voyager project, developed by a team at CMU’s MSAII program, addresses the complexities of building an adaptive travel planning system. This innovative solution harnesses artificial intelligence—including large language models (LLMs)—to generate real-time itineraries and personalized recommendations, ultimately enhancing the travel experience.

‍

Voyager’s architecture centers on a GraphDB-powered data layer, which maps relationships between travel information (e.g., destinations, traffic, events) and feeds into a Graph-RAG model layer for efficient data retrieval. A scalable serving layer, built with Next.js and Flask, integrates user-facing features, delivering an intuitive interface that accommodates dynamic user inputs.

‍

One of Voyager’s key advantages is its ability to respond seamlessly to changing travel conditions. By dynamically incorporating factors such as traffic, weather, and unexpected events, the system can adjust itineraries on the fly, providing travelers with up-to-date recommendations and insights. This adaptability significantly reduces the hassle of manual re-planning and offers a fluid decision-making process for diverse travel needs.

‍

The development of Voyager involves extensive research and system design, leveraging cutting-edge tools like LlamaIndex, Hugging Face, and Agentic, as well as various external APIs. Over the course of the project, the team refines the data pipelines, benchmarks AI models, and implements real-time updates—ensuring reliable, user-centered performance. The solution’s modular design also allows for future scalability and integration with emerging technologies.

‍

In summary, Voyager marks a forward-thinking step in AI-driven travel. By combining real-time data analysis, an adaptive pipeline architecture, and user-friendly design, the project demonstrates significant potential for reshaping how individuals plan and experience travel—both now and as new possibilities in intelligent transportation emerge.