In a major step forward for AI in energy, UBITECH’s Energy Digitalisation Group participates directly in the core development team of PowerAgent, an open-source community maintained by the Power and AI Initiative (PAI) at Harvard SEAS and dedicated to accelerating the development of LLM-powered tools and agentic AI in the power systems domain. PowerAgent forms the flagship “AI for Power” workstream within PAI’s broader research portfolio, which spans 14 projects exploring both AI for Power and Power for AI, including data centre–grid co-design and energy-aware LLM training and inference.
The community is driven by the belief that the future of agentic AI in the power domain will be shaped by the combined development and coordination of four key components: Foundation Models (PowerFM), an open-source repository of fine-tuned, domain-trained models for tasks such as load forecasting, fault detection, and grid simulation; the Model Context Protocol (PowerMCP), a collection of MCP servers that let LLMs directly interface with power system software such as PowerWorld, PSS/E, pandapower and OpenDSS; Workflows (PowerWF), which chain these tools into auditable, human-in-the-loop automation for tasks like grid impact evaluation and large-load interconnection studies; and AgentSkills (PowerSkills), which equip agents with the domain-specific knowledge and playbooks needed to sequence power system studies safely and correctly. Released under an open-source license to encourage rapid, community-driven adoption, the components are designed to work together as a coherent stack — PowerWF + PowerMCP + PowerFM — for grounding agentic AI in real operational, physics-based constraints.
By joining the core development team of PowerFM, UBITECH will actively deploy its deep domain expertise in smart grid architectures, AI applications, and data interoperability to scale the ecosystem. The company will focus on accelerating several critical phases of the foundation model lifecycle, including training, fine-tuning, deployment, and benchmarking of PowerFM against real-world operational tasks.
UBITECH’s participation underscores its commitment to driving the green transition through cutting-edge digital technology. By moving beyond generic AI models and focusing on deeply specialised, physics-informed AI for energy, the collaboration aims to bridge the gap between theoretical AI breakthroughs and the practical, real-world needs of system operators, electric utilities, consumers, and researchers alike.
Dr Magda Foti, Head of UBITECH’s Energy Digitalisation Group, commented: “Agentic AI will only be trustworthy in the power sector if it is grounded in the physics, tools, and operational discipline that grid engineers already rely on — that is exactly the philosophy behind PowerFM, PowerMCP, PowerWF and PowerSkills. Joining the PowerAgent core development team lets our Energy Digitalisation Group bring years of hands-on experience in smart grid architectures and AI-driven energy applications directly into an open-source stack that the whole community can build on, audit, and improve. It’s a natural continuation of our mission: not generic AI applied to energy, but energy-native AI built with, and for, the people who operate the grid.”
Through this collaboration, UBITECH joins a growing international community of researchers, developers, and power system experts working to establish open, interoperable, and rigorously benchmarked foundations for agentic AI in critical energy infrastructure.
