sarthak-chaturvedi / portfolio

🤖 AI Customer Support Pipeline

An event-driven AI customer support system that automates user query processing using microservices, semantic retrieval, and LLM-based response generation.


🔗 Links


🚀 Overview

This project implements an event-driven LLM support automation pipeline where multiple services process customer queries asynchronously.

User requests flow through services responsible for intent classification, knowledge retrieval, and response generation, enabling scalable and modular AI support systems.


🧱 Tech Stack

Backend

  • Node.js
  • Apache Kafka
  • PostgreSQL (pgvector)
  • OpenAI API

Frontend

  • React.js

Infrastructure

  • Docker
  • Docker Compose
  • Microservice Architecture

⚙️ Core Features

🔄 Event-Driven Microservice Architecture

Designed a Kafka-based event pipeline where services communicate through topics to process support requests asynchronously across multiple stages.


🧠 Retrieval-Augmented Response System

Integrated pgvector-based semantic retrieval to search internal knowledge base documents before generating LLM responses, improving response relevance and grounding.


📡 Kafka-Based Service Communication

Implemented Kafka producers and consumers across services using topics such as:

  • user_queries
  • classified_queries
  • retrieved_context
  • generated_responses

This enables decoupled and scalable inter-service communication.


🐳 Containerized Distributed Deployment

Containerized the entire system using Docker Compose, orchestrating:

  • Kafka brokers
  • PostgreSQL database
  • multiple Node.js microservices

for reproducible deployments and scalable processing.


🎯 Key Learnings

  • Designing event-driven distributed systems
  • Implementing Kafka-based service orchestration
  • Building LLM-powered support automation pipelines
  • Deploying multi-service architectures using Docker Compose