Brief
Global itinerary assistant with Rasa, GPT, and RAG (NDA).
Confidential client · Anonymized case study.
An AI-powered travel chatbot that searches and recommends itineraries to users worldwide. Built on Rasa NLU with LLM enhancements, the bot delivers personalized travel assistance and integrates across multiple client platforms.
Problem
Travel companies need intelligent, always-available assistants that can understand natural language queries, search itineraries globally, and deliver personalized recommendations using client-specific data.
Features
Outcome
Intent classification accuracy
0
RAG response relevance improvement
0
Chatbot uptime
0
Growth signal
Figures animate in as you scroll — pulled from this project's published results.
- Delivered a production travel chatbot with global itinerary search
- Enabled personalized recommendations through RAG and client data integration
- Successfully deployed across multiple platforms for the travel client
Challenges & solutions
- Generic vs. personalized responses → RAG with Pinecone-indexed client travel data
- Multi-platform deployment → Unified Flask backend with platform-specific integration adapters
- Complex itinerary queries → Combined Rasa NLU entity extraction with LangChain + GPT reasoning
- Context retention across turns → Redis-backed session state and conversation history
