Polyhistor - AI-Powered Travel Planning Platform
As Founding Engineer, architected a B2B2C travel platform that reduces user planning time by 40% using LLM-driven itinerary generation. Built offline-first sync layer ensuring 100% data integrity in zero-connectivity areas and cut battery consumption by 40% with custom native location engine.
Visit Polyhistor
Project Goal
Build a production-grade travel planning platform that leverages AI to personalize itineraries while maintaining flawless offline functionality and optimal battery performance.
Project Highlights
Performance Metrics
Key Features
- ▹Reduced Planning Time by 40%: LLM-driven itinerary generation creates personalized travel plans in seconds.
- ▹100% Data Integrity Offline: Architected offline-first sync layer with conflict resolution for zero-connectivity areas.
- ▹40% Battery Reduction: Custom native location engine with intelligent sampling and background optimization.
- ▹Scaled to 2,000+ Concurrent Users: Event-driven architecture with Apache Kafka for async processing.
- ▹Production AI Agents: Built multi-agent system for itinerary optimization, budget tracking, and real-time recommendations.
Technology Deep Dive
React Native (Fabric)
Built high-performance mobile app with custom native modules for location tracking and offline sync.
Node.js & Apache Kafka
Event-driven architecture handling 2,000+ concurrent users with async processing for LLM requests.
PostgreSQL & AWS
Scalable relational database on AWS RDS with read replicas for high availability.
LLM Integration
Integrated multiple LLM providers for personalized itinerary generation and natural language queries.