Prism AI: An Autonomous Multi-Cloud Diagnostic Assistant for Distributed DevOps Environments

This paper presented Prism AI, an autonomous multi-cloud diagnostic assistant that transforms DevOps incident investiga-tion from a fragmented, manual process into a streamlined, AI-driven workflow. The system integrates real-time data collec-tion from Kubernetes, AWS CloudWatch, and GitHub Actions through the Model Context Protocol, with a UnifiedAgent orchestration layer and Llama-3.3-70B inference via Groq for root-cause analysis generation. Experimental evaluation demonstrated that Prism AI re-duces investigation time from 25–45 minutes to approximately 30 seconds, achieves 92.2% diagnostic accuracy across five incident categories (compared to 78% for manual diagnosis), and eliminates context switching entirely. The system showed particular strength in identifying cascading failures that span cloud boundaries—a class of incidents especially difficult for engineers to trace manually. The modular, MCP-based architecture ensures that Prism AI can be extended to support additional cloud providers and infrastructure components, making it a scalable solution for organizations operating in increasingly complex multi-cloud environments. Future work will focus on three key directions: (i) extending MCP client support to additional cloud providers including Google Cloud Platform and Microsoft Azure, (ii) integrating predictive diagnostics for proactive anomaly detection before incidents escalate, and (iii) implementing local LLM inference fallback to eliminate external API dependency for latency-sensitive or air-gapped deployment environments.

[1]     Anthropic, “Model Context Protocol Specification,” 2024. [2]     Meta AI, “Llama 3.3: Open Foundation Models,” 2024. [3]     Groq, Inc., “Groq LPU Inference Engine Documentation,” 2024. [4]     Kubernetes Authors, “Kubernetes API Reference,” https://kubernetes.io/ docs/reference/, 2024. [5]     Amazon Web Services, “Amazon CloudWatch User Guide,” AWS Doc-umentation, 2024. [6]     GitHub, “GitHub Actions Documentation,” https://docs.github.com/en/ actions, 2024. [7]     MongoDB, Inc., “MongoDB Manual,” https://www.mongodb.com/docs/ manual/, 2024. [8]     S. Tiangolo, “FastAPI Documentation,” https://fastapi.tiangolo.com/, 2024. [9]     R. Dang et al., “AIOps: Real-World Challenges and Research Innova-tions,” IEEE Internet Computing, vol. 27, no. 3, pp. 43–51, 2023. [10]    A. Beyer et al., “Site Reliability Engineering: Monitoring and Alerting in Distributed Systems,” IEEE Software, vol. 40, no. 2, pp. 55–63, 2023. [11]    W. Zhao et al., “A Survey of Large Language Models,” IEEE Transac-tions on Knowledge and Data Engineering, vol. 36, no. 4, pp. 1892–1910, 2024.