Cleon – AI Voice Ordering Platform

Designed and developed a multi-tenant AI voice ordering SaaS enabling restaurants to accept phone orders via conversational voice agents.

• Restaurants lose revenue when calls are missed or staff are busy; phone ordering is slow, inconsistent, and error-prone.
• Integrations are fragmented (menus, merchants, voice providers), making it hard to scale across multiple restaurants.

• Designed multi-tenant architecture with strict scoping (tenant → restaurant → agent) and safe request validation.
• Implemented deterministic voice tools: menu ingestion → cart state → checkout, with fallbacks for partial failures.
• Built an integration-first admin UI so configuration lives in one place (restaurants, agents, OrderOut fields).

• Next.js (TS) admin dashboard + FastAPI backend + PostgreSQL (SQLAlchemy/Alembic).
• Voice layer: Vapi + Deepgram (STT) + ElevenLabs (TTS) with tool-call based flows.
• OrderOut integration: restaurant/menu/merchant IDs stored per restaurant; foundation for create-quote/create-order.

• Maintaining tenant isolation and preventing cross-restaurant data leaks.
• Handling unreliable external APIs during live calls (timeouts, partial data, retries).
• Keeping conversation flow stable while still allowing flexible integrations/providers.

• Working admin + integrations foundation; restaurants can be configured with agents and provider IDs.
• Clear domain boundaries enabling future provider swaps and scaling to more restaurants/tenants.

• Built a FastAPI backend with strict tenant isolation, restaurant ↔ voice-agent mapping, and scoped request validation.
• Implemented real-time AI voice agents using Vapi, Deepgram (STT), and ElevenLabs (TTS).
• Developed menu ingestion, cart management, and checkout workflows optimized for voice-driven conversations.
• Integrated OrderOut APIs for menu sync, merchant configuration, and order creation foundation.
• Created a Next.js admin dashboard to manage restaurants, voice agents, and integrations.
• Architected the system to be provider-agnostic so voice vendors can be swapped without breaking core business logic.
• Implemented SQLAlchemy + Alembic migrations for scalable schema evolution and data integrity.
• Kept clean domain separation (Tenants, Restaurants, Agents, Orders) for long-term scalability.

Next.js • FastAPI • PostgreSQL • SQLAlchemy • Alembic • TypeScript • Voice AI • Vapi • Deepgram • ElevenLabs • OrderOut • Multi-tenant SaaS