TripBooka - Full-Stack Travel Platform with AI Orchestration

TripBooka is a production-grade full-stack travel platform I built as a full-stack developer using Next.js on the frontend and a Node.js/Express backend with MongoDB. The system unifies traveler requests, agent proposal workflows, offer management, messaging, and booking operations in one product. Authentication and user lifecycle flows were integrated with Clerk, and deployment/infrastructure workflows were handled on Railway to support a reliable team delivery pipeline.

A core part of the platform is its AI-assisted proposal and planning engine powered by OpenAI. We used AI to generate key proposal content, including introduction titles, travel component suggestions, and option-level package structures, then mapped model output into normalized structured travel data. The orchestration flow resolves locations with function-calling, extracts travel intent as strict JSON, validates outputs, and auto-corrects low-confidence fields before results are returned to users and internal operators.

TripBooka also includes Scout AI, a full RAG-based matching workflow that builds embeddings for traveler requests and supplier profiles, applies filtered vector search, computes quantitative fit scores, and uses GPT ranking to recommend best-fit suppliers with contextual follow-up prompts. To keep AI quality stable in production, I helped implement confidence scoring, fallback paths for weak matches, anti-hallucination prompt constraints, rate limiting with Turnstile, and analytics for token usage, model quality, and operational performance tracking.