Archi: Agentic Operations at the CMS Experiment

2026-06-03 · Source: Takara TLDR - Daily AI Papers · Field: Science & Research — Artificial Intelligence & Machine Learning, Engineering & Applied Sciences · Depth: Advanced, medium

Summary

Archi is an open-source, end-to-end framework designed for scientific collaborations, integrating systematic ingestion and organization of diverse data sources with the deployment of configurable, private, and extensible agents. These agents retrieve and reason over the ingested data. An instance of Archi has been operational since February 2026 for the Computing Operations team of the CMS experiment at CERN's Large Hadron Collider, functioning as a support agent for technical operators. It provides retrieval and analysis capabilities by combining documentation, historical data, and live monitoring systems. Evaluations based on operator feedback and production usage queries, graded by human and automated panels, demonstrate Archi's effectiveness in resolving real-world operational tasks. The system also highlights that locally-hosted, open-weight models can perform competitively, ensuring fully private management of sensitive data.

Key takeaway

For AI Engineers developing operational support systems in sensitive environments, Archi demonstrates a viable path for deploying agentic frameworks. You should consider integrating heterogeneous data sources and utilizing locally-hosted, open-weight models to ensure data privacy and competitive performance. This approach allows for effective resolution of real-world queries while maintaining control over sensitive scientific or proprietary information, reducing reliance on external services.

Key insights

The Archi framework enables private, agentic data retrieval and reasoning for scientific operations using open-weight models.

Principles

• Combine heterogeneous data for agent reasoning.
• Prioritize data privacy with local models.
• Agentic systems enhance operational task resolution.

Method

Archi systematically ingests and organizes heterogeneous data, then deploys configurable, private, and extensible agents to retrieve and reason over this data, supporting operational tasks.

In practice

• Deploy agents for technical operator support.
• Integrate documentation, historical, and live data.
• Utilize open-weight models for sensitive data.

Topics

• Agentic AI
• Scientific Computing Operations
• CERN CMS Experiment
• Open-source Frameworks
• Data Privacy
• Open-weight Models

Best for: AI Scientist, MLOps Engineer, AI Engineer, Research Scientist

Related on AIssential

• Agent Operating Systems (AOS): Integrating Agentic Control Planes into, and Beyond, Traditional Operating Systems — Agent Operating Systems (AOS) integrate agentic control planes into OS to manage probabilistic, goal-directed AI workloads securely and accountably.
• TADI: Tool-Augmented Drilling Intelligence via Agentic LLM Orchestration over Heterogeneous Wellsite Data — Domain-specialized tool design, not just model scale, drives analytical quality in technical operations.
• MCP Marketplace Brings Real-Time Intelligence to Agentic Applications — Agentic applications require real-time external intelligence to move beyond historical data and make informed, autonomous decisions.
• Nemotron Labs: What OpenClaw Agents Mean for Every Organization — Autonomous AI agents running persistently and locally offer significant advantages for continuous, high-iteration tasks.
• AI Models as a Service: Powering Agentic AI, Privacy, & RAG — Models as a Service provides a sovereign, internal API gateway for managing and deploying multiple AI models.
• Oracle launches agentic AI tools for enterprise data management — Oracle's expanded AI Database integrates agentic AI directly with enterprise data, enhancing security, privacy, and performance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.