Ahmed Al Jawad Shahir is a MEng student in the Human-Centered Machine Intelligence Lab, working on AI-Powered Reliability and Exposure Auditing for MCP-Enabled Agentic Systems. His research develops a trace-based auditing and evaluation framework for systems built on the Model Context Protocol (MCP), addressing reliability and exposure risks—such as cross-context aggregation, inference amplification, reasoning instability, and limited auditability—that emerge at the orchestration layer when LLM agents are connected to enterprise tools, retrieval systems, and structured data. The work aims to deliver a reusable evaluation pipeline, measurable reliability metrics, and empirical, client-facing guidance for deploying MCP-based AI systems more confidently.