Published Paper Details:

Modern distributed systems demand proactive and intelligent incident management to maintain service reliability and reduce mean time to recovery (MTTR). This paper presents AgileCopilot, a proof-of-concept AI-powered Site Reliability Engineering (SRE) module that integrates Machine Learning (ML) anomaly detection with Generative AI root cause analysis (RCA). The system comprises four components: a Victim Service simulating a live web application with controllable fault injection, an ML-based Watchdog using the Isolation Forest algorithm, an AI Investigator powered by Gemini Generative AI, and a real-time Streamlit monitoring dashboard. The Watchdog monitors CPU utilization, memory consumption, and HTTP response latency; upon detecting an anomaly validated by a heuristic intensity filter, it triggers the AI Investigator to produce a structured JSON report containing the root cause and remediation steps. Experimental results demonstrate accurate detection of injected anomalies while suppressing false positives during normal operation, validating the effectiveness of combining unsupervised ML with rule-based filtering. AgileCopilot demonstrates the viability of embedding autonomous SRE intelligence into software engineering workflows.