Justin Arndt

-

Bio

Independent AI safety engineer & systems architect building mathematically verified evaluation harnesses, MCP cryptographic security tooling, and autonomous agent containment suites.

Projects

Autonomous Agent Sandbox Escape & Containment Benchmark

creator

Detecting tool-poisoning and schema drift in MCP servers

creator

Fixing flaky AI benchmark scorers with eval-invariance-engine

creator

Grants

No grants recorded.