Sage Grande

An Open Artificial Intelligence Testbed for Edge Computing and Intelligent Sensing

Artificial Intelligence for Real-Time Edge Computing and Sensing

Sage Testbed Nodes, used for Development and Testing

The Sage Grande Testbed (SGT) is building a cutting-edge artificial intelligence (AI) cyberinfrastructure to support advanced AI research.

SGT, funded by the NSF Office of Advanced Cyberinfrastructure, provides access to AI-enabled edge computing resources and software tools integrated with sensors—including infrared and RGB cameras, microphones, and a variety of atmospheric and air quality instruments—deployed across natural, urban, and wildfire-prone environments, with networking capabilities that support real-time hazard reporting.

By bringing advanced AI to the edge, where data is collected, full-resolution analysis, dynamic automation, and immediate actionable responses can be computed. Each Sage node includes a GPU and AI-optimized software stack connected to instruments such as infrared cameras, RGB cameras, LiDAR, and traditional sensors for air quality and wind, as well as LoRaWAN connected sensors for low-bandwidth measurements such as soil moisture. With over 100 Sage nodes deployed across 17 states, SGT provides a national-scale testbed for AI-enabled, autonomous, and rapid-response science and sustained observation of ecological systems, agriculture, urban environments, and weather-related hazards.

  • 2019: Sage is Founded
    • NSF MSRI-1 Project
    • In 2019, as an NSF MSRI-1 project (#1935984), Sage built the nation’s first distributed, edge computing resource for AI and intelligent sensing.
  • 2023: AI Cyberinfrastructure Exploration
    • NAIRR Pilot Demonstration
    • In 2023, Sage began exploring how a national-scale testbed could provide AI as an open research resource (#2331263) and became a NAIRR Pilot Demonstration.
  • 2025: A National AI Testbed
    • Sage Grande Testbed (SGT)
    • In early 2025, NSF funded the Sage Grande Testbed (SGT) as a five-year project (#2436842) to build out the Sage infrastructure for artificial intelligence, the integration of advanced large language models (LLMs) for the edge, and new end-to-end workflows.

Existing Sage nodes have been integrated into major scientific infrastructure networks such as NEON, OHAZ, Hawaiʻi Mesonet, and HPWREN, enabling researchers to conduct convergent research across disciplines. The Sage ecosystem includes a flexible software stack for edge computing, a Sage Portal for real-time data monitoring and edge AI job management and the Edge Code Repository for deploying reusable AI workflows. Sage nodes already support multimodal sensor integration, data streaming, AI inferencing, and cloud-archiving of results for scientific reuse.

High-level overview of software infrastructure

Sage Grande will extend this infrastructure with more powerful hardware (NVIDIA Orin NX and AGX modules), additional high-bandwidth and high-resolution sensors, expanded software tools, including SageChat, which provides interfaces to LLMs running at the edge, and advanced edge-to-cloud orchestration, data commons integration, and support for AI model evaluation and privacy-enhancing technologies.

SGT integrates advanced edge computing, multi-modal sensing, and LLMs into an accessible, user-friendly research testbed. The platform allows scientists and students to deploy AI applications directly in the field, conduct high-resolution measurements of environmental conditions, and interact with complex datasets using natural language prompts. Sage Grande bridges the gap between laboratory-scale AI development and real-world, mission-critical deployment scenarios across fields such as urban science, agriculture, ecology, fire science, and disaster response.

Research Team

Current Members

Learn More

Funding

This work funded in part by a National Science Foundation grant — Sage Grande: An Open Artificial Intelligence Testbed for Edge Computing and Intelligent Sensing. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.