Back to Projects

Telecom KPI Anomaly Detection Using AI Agents

LangGraph, MCP, LLMs — Multi-Agent AI System for Real-Time KPI Monitoring

Telecom KPI Anomaly Detection

📋 Project Overview

Architected a multi-agent AI system integrating LangGraph and MCP (Model Context Protocol) with anomaly detection models (DWT-MLEAD, Isolation Forest, Ensemble Voting) and NVIDIA LLMs to track more than 10 KPIs across 100+ cellular sites in real time.

Elevated anomaly detection reliability by 30% using ensemble modeling and automated hypothesis validation. Deployed a Dockerized conversational analytics platform with Gradio UI, enabling engineers to query system health, KPI correlations, and root causes in natural language.

⚡ Key Highlights

  • Multi-Agent Architecture: LangGraph and MCP orchestration for coordinated AI workflows
  • Anomaly Detection: DWT-MLEAD, Isolation Forest, and Ensemble Voting models
  • Scale: 10+ KPIs monitored across 100+ cellular sites in real time
  • 30% Reliability Improvement: Ensemble modeling and automated hypothesis validation
  • Conversational Analytics: Gradio UI for natural language queries on system health, KPI correlations, and root causes
  • Deployment: Dockerized platform for production-ready analytics

Skills Demonstrated

LangGraph MCP LLMs NVIDIA Anomaly Detection Isolation Forest Ensemble Learning Gradio Docker Python

More details and images coming soon.