VENUEPLACE
Marketplace for booking event venues — think Airbnb for events, real-time availability, instant quotes, and built-in payments. As CTO, I designed the architecture and built the entire platform end-to-end.
- Role
- Everything technical, solo — architecture to deployment
- Timeline
- Prototype, then a from-scratch Next.js rebuild, then the first live venues about two months later
- Live
- venueplace.com · English, Estonian, Latvian
- Inventory
- 100+ published venues across Estonia
Event planners in Estonia had to email venues one by one to learn prices and capacities, and hosts had no simple way to take bookings online. Planning a party, a seminar or a wedding meant a spreadsheet of unanswered emails.
- A two-sided marketplace. Hosts publish spaces, packages, catering and accommodation with transparent pricing; guests compare real prices, capacities and photos and book directly, with Stripe payments, localized invoices and SEPA bank transfer.
- The whole platform end to end: a Next.js codebase with its own API and database, sign-in, Google Calendar two-way sync and iCal feeds, three interface languages, automated test suites and scheduled jobs.
- An AI assistant that turns a plain-language brief into a booking: it asks structured follow-ups, searches the live inventory including availability, builds a comparison shortlist, answers detail questions by quoting venue text and hands the user a pre-configured booking page. Answers are grounded only in real data — no invented venues, no invented prices.
- An AI translation layer: host-written venue content is translated into every interface language, stored per language and regenerated whenever the host edits.
- A data-driven SEO layer: keyword landing pages (city × event type) filled from live inventory and only opened to search engines once enough venues actually match, around a hundred venue pages in three languages, structured data throughout, sitemaps with honest last-modified dates, llms.txt.
- An internal admin app, plus a secured MCP server that lets AI clients read platform data — read-only, nothing can be changed through it.
From an AI-generated prototype to a from-scratch Next.js rebuild to the first live venues — about two months from rebuild to launch. AI translation, the AI assistant and the SEO layer all shipped within the first eight months.
Since the landing pages went live, search impressions have climbed steadily and active users are up by more than half against the previous month, new users by more than 80%.
- Next.js 15 App Router with ISR, React 19, TypeScript, Tailwind 4, shadcn/ui, next-intl
- tRPC 11 · Prisma 7 on Supabase Postgres · Supabase Auth
- Stripe PaymentElement, webhooks, localized invoices, platform merchant-of-record
- Vercel AI SDK with Claude and OpenAI models, prompt caching, per-user rate and token caps
- Google Calendar sync, iCal feeds, Vercel crons, Playwright + Vitest