Voyagenie

B2CBeta

Your Trip. Perfectly Planned.

AI-powered travel itineraries that actually understand what you want—from neighborhood vibes to realistic budgets, generated in seconds.

GeminiTravelMulti-stage LLMReact/TS

Launch App

The Problem

Trip planning is a time sink. You spend hours juggling Google Maps, travel blogs, TikTok recommendations, and budget spreadsheets—only to end up with a disjointed list that doesn't account for travel time, local timing, or your actual preferences.

The Approach

Voyagenie treats itinerary generation as a multi-stage intelligence problem, not a single prompt. The system builds destination intelligence, architects trip structure based on pace and interests, generates activities with realistic timing, then enriches everything in parallel—social proof hashtags, budget breakdowns, and pre-trip todos.

Features

5-step preference wizard capturing destination, interests, budget, and pace

Multi-day itinerary generation with realistic timing and neighborhood-aware routing

Detailed budget breakdown across 6 categories with per-day costs

Social proof hashtags linking to TikTok, Instagram, and YouTube searches

Natural language refinement ('add more cafes', 'swap Day 2 morning')

One-click HTML export for printing or sharing

Technical Highlights

7-stage LLM chain with parallel execution

Schema-validated outputs with TypeScript types matching Gemini definitions

Intelligent refinement with confidence threshold triggering clarification

Caching destination intelligence reduces API costs ~40%

Learnings

Chain decomposition beats prompt engineering—mediocre prompts at each stage outperform a brilliant single prompt

Schema validation is non-negotiable for LLM apps—eliminated 15% parsing failures

Parallel execution matters at scale—dropped p95 generation time from 45s to 18s