AI marketplace·2026

NativeTravel — an AI-first travel marketplace

A two-sided marketplace of local experiences with AI personalization: tours, restaurants, events and activities. A multi-tenant architecture supporting multiple cities and brands, semantic search on pgvector, AI recommendations, translations and moderation via Claude. A Fastify + tRPC backend and a Next.js 15 + React 19 frontend. The project is in active development.

FastifytRPCPrismaPostgreSQLpgvectorNext.jsReact 19Claude
NativeTravel — an AI-first travel marketplace

Project goals

  • Build a two-sided marketplace of local experiences with AI matching
  • Implement multi-tenancy for multiple cities and brands
  • Introduce semantic search and recommendations on embeddings
  • Keep the app working even without AI keys (deterministic fallback)

Tasks

  • Design a monorepo: a Fastify + tRPC v11 backend and a Next.js 15 frontend
  • Model the domain (~45 Prisma models, 27+ modules)
  • Wire up PostgreSQL 15 + pgvector for embeddings and semantic search
  • Integrate Claude (chat, translations, moderation, vision) and OpenAI (embeddings)
  • Implement bookings, payments (Stripe/Kaspi) and host payouts
  • Build a partner REST API for channels (OTA) with Bearer auth
  • Set up deployment via Docker Compose behind a Caddy reverse proxy

Results

Measurable performance indicators

  • Ready infrastructure: monorepo, design system, i18n and an AI layer
  • Semantic search and recommendations on pgvector embeddings
  • Multi-tenancy: multiple cities, custom domains, per-tenant AI budgets
  • A partner REST API for OTAs and channel managers
  • A deterministic stub mode — the app runs without AI keys

Key metrics

Specific performance indicators of the project

45+

Data models

27+

Backend modules

3

Languages

pgvector

Semantic search