AI/ML Model Cards in Edge AI Cyberinfrastructure: towards Agentic AI

2025-11-26

AI Summary

Static model cards in edge AI fail to support effective lifecycle assessment due to their inability to adapt dynamically during deployment. Method: This paper reconceptualizes model cards as dynamic, ecosystem-embedded objects within the ICICLE AI framework. Its core innovation is the Model Context Protocol (MCP)—a lightweight, context-aware protocol enabling proactive dialogue and runtime interaction, replacing conventional RESTful interfaces to achieve service-oriented model cards, enhanced interpretability, and runtime governance. Results: Experiments demonstrate that MCP incurs only controllable performance overhead in edge environments while significantly improving dynamic update capability, contextual adaptability, and collaborative potential with agent-based AI systems. MCP thus establishes a practical, deployable technical paradigm for end-to-end AI lifecycle governance.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Multiagent Systems: Modeling other Agents
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

AI/ML model cards can contain a benchmarked evaluation of an AI/ML model against intended use but a one time assessment during model training does not get at how and where a model is actually used over its lifetime. Through Patra Model Cards embedded in the ICICLE AI Institute software ecosystem we study model cards as dynamic objects. The study reported here assesses the benefits and tradeoffs of adopting the Model Context Protocol (MCP) as an interface to the Patra Model Card server. Quantitative assessment shows the overhead of MCP as compared to a REST interface. The core question however is of active sessions enabled by MCP; this is a qualitative question of fit and use in the context of dynamic model cards that we address as well.

Problem

• Dynamic model cards track AI usage over time
• Assess MCP interface benefits for model card servers
• Evaluate active sessions enabled by MCP protocol

Innovation

• Dynamic model cards track lifetime usage patterns
• MCP interface enables active sessions for model cards
• Quantitative assessment compares MCP overhead to REST