Pathfinder

An AI-powered travel planning application that transforms natural-language requests into structured, location-aware itineraries.

Stay tuned:)

Overview

Pathfinder is a full-stack travel planning app built during a 24-hour hackathon. It allows users to describe travel plans in natural language and instantly receive a structured, editable itinerary. The system combines large language models with real-time mapping data to produce location-aware recommendations and optimized routes.

The project focuses on fast backend orchestration, AI integration, and turning unstructured user input into usable planning data.

Key Features

  • •Natural-language travel planning (e.g., “3 days in Tokyo with food and museums”)
  • •AI-generated, structured itineraries with editable sections
  • •Location-aware recommendations powered by live map data
  • •Intelligent place suggestions and routing based on user intent
  • •End-to-end flow from free-text input to actionable travel plans

Tech Stack

Frontend

React, TypeScript, Vite

Backend

Python, FastAPI, Pydantic, OAuth2, JWT

Database

SQLite

DevOps

Docker, Nginx

AI & APIs

Jest, React Testing Library

Challenges & Solutions

Challenge:

Parsing vague or ambiguous natural-language travel requests into structured itinerary data

Solution:

Designed custom prompt structures and backend orchestration to consistently transform free-text input into well-defined itinerary objects.

Challenge:

Ensuring recommendations are relevant to real-world locations and distances

Solution:

Combined LLM-generated plans with real-time Google Maps data to validate locations, calculate routes, and refine suggestions.

Challenge:

Delivering a complete product within a 24-hour hackathon timeframe

Solution:

Prioritized backend logic and core user flow, enabling a functional end-to-end experience with minimal UI overhead.

Learning Outcomes

  • ✓Gained experience translating natural-language user input into structured, actionable data using large language models.
  • ✓Learned how to combine LLM-generated outputs with real-time external APIs (Google Maps) to produce location-aware, practical results.
  • ✓Strengthened backend orchestration skills by designing workflows that transform free-text requests into editable travel itineraries.
  • ✓Developed an understanding of prompt design and constraint enforcement to keep AI-generated plans consistent and usable.
  • ✓Improved rapid prototyping skills by delivering a complete, end-to-end product within a 24-hour hackathon timeframe.