Alignment & Safety Researcher

The Role

This is not a PR role. Ceridwen.ai's safety mechanisms are legally required by the MABOS license. Attestation integrity, stability lockouts, and governance compliance are not features — they are architectural invariants. You ensure they hold.

What You'll Do

• Design and validate safety mechanisms across the MABOS cognitive architecture
• Extend and maintain the attestation system that ensures architectural integrity
• Research adversarial failure modes and develop countermeasures
• Collaborate with governance and legal teams to ensure safety mechanisms meet regulatory requirements
• Build evaluation frameworks that catch the kind of failures other companies discover after three months of hallucinated data

Requirements

• PhD or equivalent research depth in AI safety, alignment, formal verification, or related fields
• Published work in mechanistic interpretability, alignment, robustness, or verification
• Systems engineering capability — you can implement your research in production code, not just papers
• Failure mode expertise. Deep understanding of failure modes in autonomous systems
• You believe safety is an engineering discipline, not a philosophy department

Compensation

All positions include equity in Ceridwen.ai. Salaries are TBD and will be determined based on role scope, experience, and what you bring to the table. We will not insult you with a lowball offer, and we expect you not to waste our time with inflated expectations disconnected from contribution.

The Builder Clause

We don't care where you went to school. We don't care if you went to school. Our founder is self-taught, started coding at 13, and built a 602,000-line cognitive architecture without a CS degree.

Meet the qualifications, or show us what you've built. Either path works. Both paths demand excellence.

Apply

Ready to move? Send a short note of relevant proof-of-work — past shipped projects, metrics you moved, or a draft of your first 30 days.