Back to Projects
AI SaaS Platform
Innspeed
2026
Innspeed — AI Guest Communications Platform
Innspeed is a multi-tenant SaaS platform that helps independent and boutique hotels automate guest conversations across web chat and WhatsApp. Hotels build an AI knowledge base from their website, embed a guest widget, connect WhatsApp Business, and manage escalations, guests, campaigns, and billing from a single dashboard. The product is designed for European hotels with GDPR-minded data residency and EU AI Act–aligned audit logging.
Key Features
- Multi-tenant hotel onboarding with Stripe checkout, 14-day trials, plan caps, and subscription lifecycle management
- Embeddable Preact guest chat widget (Shadow DOM, SSE + polling, ratings, lead capture, multilingual system messages)
- Grounded AI agent (tool-calling) over vector search of hotel Q&A and scraped content chunks, with staff escalation
- Unified inbox for web + WhatsApp with real-time updates, assignment, resolve/reopen, and AI lock
- Website discovery/scrape pipeline that extracts knowledge-base entries and embeddings
- WhatsApp Business Cloud integration (Embedded Signup, templates, inbound webhooks, campaign template sends)
- Guest CRM with tags, reservations, CSV/XLSX import, and marketing consent tracking
- Scheduled and broadcast WhatsApp campaigns (pre-arrival, upsell, checkout, post-stay) with delivery tracking
- Analytics dashboard (conversation volume, automation rate, CSAT, response-time percentiles, channel/language, campaign funnel)
- Team invites with admin/staff RBAC, email OTP two-factor auth, and plan-gated settings
- Marketing site, pricing, contact form, and guided 4-step product onboarding
Technical Architecture
- Modular monolith backend: Fastify feature modules (routes → controller → service → mapper) under /api/v1
- Multi-tenant hotel isolation via JWT claims (hotel_id, app_role), Postgres RLS, and admin/staff policies
- AI agent loop with tools (intent, KB search, hotel context, opt-out, escalate, finalize) and ai_pipeline_logs
- Async scrape jobs via Redis/arq with signed webhooks back to the API; campaign tick via in-process cron or secret-gated HTTP
- Frontend feature folders with App Router route groups; dashboard bootstrap fan-out; axios silent token refresh
- Public widget API with SSE staff-reply stream; dashboard Realtime over Supabase channels
Highlights
Built an end-to-end hospitality AI platform spanning a Next.js dashboard, embeddable guest widget, Fastify API, vector knowledge base, WhatsApp automation, and Stripe SaaS billing.
Strong multi-tenant Postgres design with RLS, grounded LLM tooling on AWS Bedrock, real-time inbox operations, and production-grade webhook/idempotency patterns demonstrate full-stack systems design across product, security, and integrations.