Visito AI
- Designed and implemented production-grade AI systems for hospitality conversational agents handling hundreds to thousands of daily conversations — including RAG pipelines, multi-LLM orchestration (OpenAI + Gemini), tool calling, embeddings, and advanced prompting strategies.
- Built the full conversation escalation flow, including real-time push notifications to route frustrated users to human agents, and developed a frustration-detection module that scored conversation sentiment periodically to proactively surface at-risk interactions.
- Architected and scaled the core backend platform using Node.js and TypeScript, owning REST APIs, webhook integrations (Meta, Stripe), and real-time messaging workflows.
- Rebuilt the real-time communication layer into an event-driven architecture using RabbitMQ and WebSockets, improving reliability and throughput under concurrent traffic.
- Led a full cloud migration from Azure to GCP — moving all RabbitMQ consumers, webhook endpoints (Meta, Stripe), and the core API to dedicated cloud infrastructure — significantly reducing costs and eliminating the downtime issues present in the previous environment.
- Joined as the sole founding engineer, reporting directly to the CTO with full ownership of technical and architectural decisions — playing a key role in taking Visito from early stage to a successful production platform in the hospitality AI space.