An event-driven AI customer support system that automates user query processing using microservices, semantic retrieval, and LLM-based response generation.
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.
Backend
Frontend
Infrastructure
Designed a Kafka-based event pipeline where services communicate through topics to process support requests asynchronously across multiple stages.
Integrated pgvector-based semantic retrieval to search internal knowledge base documents before generating LLM responses, improving response relevance and grounding.
Implemented Kafka producers and consumers across services using topics such as:
user_queriesclassified_queriesretrieved_contextgenerated_responsesThis enables decoupled and scalable inter-service communication.
Containerized the entire system using Docker Compose, orchestrating:
for reproducible deployments and scalable processing.