Basil Korompilias

Senior Technologist operationalizing AI Safety standards (EU AI Act, NIST AI RMF, ISO 42001, UNESCO AI Ethics) into production-ready infrastructure. Leverages 12+ years of executive R&D management and 8 years of frontier safety engineering to build alignment tooling that scales from research to governance. - Developed an epistemic AI safety risks taxonomy validated on 90M+ sparse autoencoder features across 16 language models (via Neuronpedia API). - Built an Inspect AI evaluation suite with 20 metrics across 5 domains (hallucination, sycophancy, deception/goal drift diagnostics). - Shipped socio-technical specifications and a custom llama.cpp kernel to support zero-trust AI governance architectures.

Personal Website

https://gyrogovernance.com/

LinkedIn

LinkedIn

Twitter / X

-

LessWrong

-

EA Forum

-

Projects

AIR: Platform Mitigating X-Risk through Human Capacity Ops

Team Member

No grants recorded.