Agentic Chatbot

Enterprise AI agent with RAG, OCR, PostgreSQL integration, and tool orchestration deployed on AWS.

Overview

This project presents an agentic AI system designed for hotel assistance, built with modern LLM technologies and agent frameworks. The chatbot uses LangGraph to orchestrate intelligent workflows and interact with external tools such as a PostgreSQL database and a Qdrant vector database.

By leveraging Retrieval-Augmented Generation (RAG), the system can provide accurate and context-aware answers about hotel policies, services, and frequently asked questions.

The assistant is designed to function as a digital concierge, capable of retrieving hotel information, interacting with operational data, and assisting users with tasks such as reservations or service inquiries. By combining language model reasoning with structured and unstructured data sources, the system delivers reliable responses while maintaining a natural conversational experience.
The general arquitecture diagram of the proposed Agentic AI sytem

git repo here: github

Core Capabilities

1. Multi-Agents Processing:

  • NLP: Natural language understanding via OpenAI GPT-4, Gemini, Antrophic LLMs

2. RAG System:

  • Pinecone/QDrant: vector database for semantic search
  • Knowledge base: hotel policies, restaurant menus, FAQs
  • Context-aware information retrieval with metadata filtering

3. Database Integration:

  • PostgreSQL for reservations, customer data, and audit logs
  • Secure access with RBAC and parameterized queries
  • Connection pooling and query optimization

4. Tool Orchestration:

  • Create/cancel reservations
  • Check availability and process payments
  • Send confirmation emails (mcp)

System Architecture

graph TD
    User([User Query]) --> Agent[ReAct Agent]
    Agent -->|Decide Tool| Tools{Tools Selection}
    Tools -->|Policy Inquiry| RAG[tool_get_hotel_info_rag]
    Tools -->|Availability/Rooms| SQL[SQL DB Tools]
    Tools -->|Booking/Check-in| Actions[Action Tools]
    RAG --> Agent
    SQL --> Agent
    Actions --> Agent
    Agent -->|Final Answer| User

Technical Implementation

Agent Framework:

  • LangChain: Agent orchestration and tool management
  • MCP (Model Context Protocol): Standardized tool communication
  • Advanced Prompting: Re Act reasoning, few-shot examples

Deployment Architecture

Infrastructure:

  • Containerization: Docker multi-stage builds
  • Orchestration: Kubernetes on AWS ECS
  • Database: AWS RDS PostgreSQL with automatic backups
  • Monitoring: CloudWatch, LangSmith.

Use Cases

Hotel Management:

  • Automated check-in with ID verification
  • Concierge services and local recommendations
  • Reservation management and payment processing

Restaurant Operations:

  • Table reservations and waitlist management
  • Menu information and allergen queries
  • Order processing with handwritten note OCR

Technology Stack

Core Technologies:

  • Python 3.11+, LangChain, OpenAI GPT-4
  • PostgreSQL, Pinecone
  • FastAPI, Docker, Kubernetes

Infrastructure:

  • AWS (ECS, RDS, S3, lambda, CloudWatch, Secrets Manager)
  • Terraform (Infrastructure as Code)

Key Achievements

This project demonstrates senior-level expertise in:

  • ✅ Agentic AI systems with autonomous tool use
  • ✅ Production-grade cloud deployment (AWS + Kubernetes)
  • ✅ Database design with security best practices
  • ✅ Advanced prompt engineering and LLM orchestration

This project showcases the ability to design and deploy enterprise-grade AI systems. For collaboration or technical discussions, please contact me.