Sami Lafrance
Founding Engineer @ Ctrl+G
About
Building evaluation systems and security datasets for AI agents in cybersecurity. The goal: measure what models actually understand, not just what they generate.
Work
- env — CVE exploitation and patching environments
- data — Security data generation and curation
- pipe — Pipelines for creation, review, and validation
- bench — Agent benchmarking on real-world vulnerabilities
Real systems, multi-language codebases, reproducible vulnerabilities, end-to-end evaluation.