Welcome to EPRI Current – your go-to source for the latest in energy innovation, research, and insights.
Your host, Samantha Gilman, meets with EPRI experts and industry leaders to unpack the energy sector’s most pressing challenges and promising opportunities. Through timely and thought-provoking conversations, the podcast explores critical themes including advanced energy technologies, grid reliability and flexibility, climate adaptation and resilience, the transformative role of artificial intelligence, the evolving economics and policies shaping energy systems, and more.
Whether you're an energy professional, policymaker, or curious listener, the EPRI Current brings you expert perspectives that highlight how innovation and collaboration are powering a more resilient and reliable energy future.
Tune in and be part of the journey as we shape the future of energy—together.
Latest Episodes
Defining Flexibility: How Flex MOSAIC™ Creates a Shared Language for the Grid
Host: Samantha Gilman
Guests: Dave Weaver, Exelon; Clift Pompee; Compass Data Centers; Anuja Ratnayake, EPRI.
In this week’s episode of the EPRI Current, Dave Weaver from Exelon, Clift Pompee from Compass Data Centers, and Anuja Ratnayake from EPRI join host Samantha Gilman to discuss how new approaches to flexibility can accelerate grid connections amid surging demand.
Experts focus on EPRI’s recently launched Flex MOSAIC™, a framework developed through EPRI’s DCFlex initiative to define what flexibility means in practice by creating a shared, performance-based language between utilities and large load customers. By defining five distinct classes of flexibility with clear parameters such as notification time, duration, and frequency of use, Flex MOSAIC™ helps align expectations upfront and reduce the ambiguity that has historically slowed interconnection.
Tune in to hear about how concepts like “headroom” and performance-based flexibility classes can help unlock existing grid capacity, reduce interconnection timelines, and support reliability. With perspectives from across the energy ecosystem, the episode highlights both the opportunity and the coordination required to enable faster, more adaptive grid integration.
Learn more about EPRI’s DCFlex Initiative: https://dcflex.epri.com/
Learn more about EPRI’s Flex MOSAIC™: https://dcflex.epri.com/flex-mosaic
Learn more about headroom: https://www.epri.com/research/products/000000003002034162
Micro Data Centers and the Evolution of AI Infrastructure
Host: Samantha Gilman
Guests: Ben Sooter, EPRI
In this episode of the EPRI Current, host Samantha Gilman speaks with EPRI’s Director of Agentic AI and Micro Data Centers, Ben Sooter, about the growing role of distributed inference, also known as micro data centers, in the evolving AI landscape. While large, centralized data centers dominate current discussions, Sooter explains why the next wave of compute demand will shift closer to end users to support faster, latency-sensitive applications.
The conversation highlights how this distributed approach could reshape grid planning, including opportunities to leverage available capacity at distribution substations and the need to better understand emerging load profiles. Sooter also discusses EPRI’s collaborative pilot efforts with NVIDIA, Prologis, and InfraPartners to explore siting, deployment, and scalability. As AI adoption accelerates, this episode examines how utilities can prepare for and help enable the next phase of digital infrastructure.
Beyond the Benchmark: Evaluating AI for Real World Use
Host: Samantha Gilman
Guests: Jaime Sevilla, Epoch AI; Apurba Sakti, EPRI.
How should organizations interpret AI benchmarks – and where do they fall short when moving from pilots to real world deployment?
In this episode of EPRI Current, host Samantha Gilman is joined by Jaime Sevilla, Director of Epoch AI, and Apurba Sakti, EPRI Principal Technical Leader for AI, for a deep dive into AI benchmarking and responsible adoption. The conversation explores why strong benchmark scores don’t always translate into operational readiness, the limitations of generic leaderboards, and why domain and workflow-specific evaluations are critical – especially in high-consequence sectors like energy. The discussion highlights how organizations can move beyond demonstrations toward continuous, evidence based evaluation to ensure AI systems are reliable, transparent, and fit for real world use.
