Shane McFly

For twelve years I've worked on the security of systems that can't be allowed to fail. At the National Laboratory of the Rockies (formerly NREL, 2019–2026) I was Senior Cybersecurity Researcher and Principal Investigator: I directed the network/communications pillar of Liberty Eclipse, DOE CESER's full-scale national grid-recovery exercise; engineered the ARIES Cyber Range into a high-fidelity emulation of multi-owner power systems; and led the lab's initiative integrating Cyber-Informed Engineering into national lifecycles. Earlier, I built smart-grid black-start simulations for DARPA RADICS and assessed embedded aircraft network vulnerabilities at Boeing Research & Technology. Alongside that work I founded and chaired the DEFCON Ethics Village (2017–2019), have presented at DEFCON's ICS Village for the last three years, and teach graduate-level system security as adjunct faculty at Colorado School of Mines. I co-authored a 2019 paper "A Graph-Based Analysis of Industrial Control Systems Network Traffic" Imani Palmer, Shane McFly and Edmond Rogers. In May 2026 I left the lab to move into AI safety full-time. The same engineering discipline that keeps the grid trustworthy — threat modeling, engineered-in consequence limits, verifiable state — is missing from how AI agents handle memory and persistent context. I've been working on this already: a working paper on context persistence and state integrity in distributed multi-agent LLM architectures, and Gemma/Artoo, a multi-agent system I built and operate on my own hardware as a testbed for auditable agent memory.

LinkedIn

LinkedIn

Projects

No grants recorded.