Overview
Designed and built an end-to-end AI-powered hotel search assistant from scratch.
The system understands natural language requests, extracts user constraints, retrieves relevant hotels, and generates grounded responses based on real search results.
Key Features
- Natural language understanding
- Retrieval-Augmented Generation (RAG)
- Vector search using embeddings
- Constraint matching and filtering
- Structured hotel retrieval pipeline
- Production-oriented architecture
- End2end evaluation framework
Technologies
- Python
- FastAPI
- OpenAI API
- Elasticsearch
- Vector Search
- Pydantic
- AsyncIO
My Responsibilities
- Designed the overall system architecture
- Implemented retrieval and ranking pipeline
- Built hotel embedding pipeline
- Developed structured data schemas
- Implemented evaluation framework for retrieval and end2end quality
- Integrated LLM reasoning with deterministic retrieval
- Built API and backend logic
GitHub:
https://www.twine.net/signin
- AI Engineer