Self-Hosted Travel Platform with Sovereign Geographic Data

Most travel apps optimize for speed—shortest routes, popular landmarks, predictable paths. Haydov takes the opposite approach: enabling slow travel by surfacing meaningful detours and lesser-known places along the way.

I designed and built Haydov as a fully self-hosted travel platform powered by OpenStreetMap, with an AI layer that suggests contextual deviations from a planned route—encouraging exploration beyond predefined itineraries.

The system is built on sovereign geographic data: routing, geocoding, and indexing are all owned end-to-end, with no dependency on commercial APIs.

Core Capabilities

• Self-hosted routing and geocoding (Valhalla + Pelias)
• AI-driven suggestions to enrich travel routes with nearby points of interest
• Full geographic data pipeline from raw extracts to production services
• Mobile app for route visualization and interaction
• Extensible provider system for adding new geographic data sources

Key Results

• Built a complete mapping stack with no reliance on commercial APIs
• Enabled automated ingestion and indexing of global geographic datasets
• Designed a modular pipeline supporting multiple data providers via plugins
• Delivered a mobile-first experience for exploratory travel

Architecture Overview

Haydov is a polyglot distributed system spanning Rust, Node.js, Go, and mobile clients:

• Data Pipeline (Rust + WebAssembly)
  Downloads and processes geographic data using pluggable provider components loaded at runtime
• Import & Indexing (Node.js + RabbitMQ + gRPC)
  Coordinates ingestion into a self-hosted Pelias geocoding stack backed by Elasticsearch
• Routing Engine (Valhalla)
  Provides route computation using the same OSM-derived data
• Mobile App (React Native / Expo)
  Displays routes and integrates AI-driven suggestions for exploration

Technical Highlights

• Designed a WebAssembly plugin system (WIT interface) for geographic data providers
• Built a streaming ingestion pipeline with S3-compatible storage and multipart uploads
• Implemented event-driven orchestration via RabbitMQ
• Deployed full stack on Kubernetes with reproducible environments (Nix + Tilt + Kustomize)
• Developed a Go CLI with terminal UI for managing local environments
• Implemented polyline encoding/decoding for efficient route serialization

My Role

Owned the system end-to-end:

• Designed architecture and data pipeline
• Built core services (Rust downloader, Node.js importer)
• Implemented orchestration and deployment setup
• Developed mobile application and developer tooling