Founding Engineer Infrastructure
French nonprofit research organization that develops AI risk measurement tools, evaluates frontier-AI company risk-management practices, and contributes to AI standards and policy.
Funding history
About SaferAI
SaferAI is an independent French nonprofit association focused on reducing risks from advanced AI. Its work combines quantitative AI-risk research, public frontier-company risk-management ratings, and standards and governance work for governments and industry.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will maintain and improve cloud infrastructure for model evaluations while protecting sensitive data. You will also support model evaluation and mitigation work by improving evaluation tooling, selecting and running benchmarks, analyzing results, and developing or implementing safety mitigations.
Requirements
- Strong AWS or equivalent cloud-provider experience, including EKS or Kubernetes, Helm, and Karpenter
- Infrastructure as Code experience with Pulumi, Terraform, or similar tools
- Python experience
- CI/CD workflow experience
- PostgreSQL operations experience, including RLS
- Security engineering fundamentals, including zero-trust access, Kubernetes network policy, AWS GuardDuty, least-privilege IAM, and secrets management
- 3+ years of professional platform, infrastructure, or DevOps experience
- Experience delivering projects in fast-paced, evolving environments
- Ability to produce clear written deliverables under time pressure and incomplete information
- Ability to work effectively in a remote, asynchronous, international environment
- Alignment with AI risk-management and safety goals
Responsibilities
- Improve and maintain cloud infrastructure for model evaluations
- Maintain cybersecurity practices that safeguard sensitive laboratory data
- Drive model evaluation and mitigation work
- Improve Inspect AI tooling for evaluation needs
- Identify and run risk-relevant benchmarks on AI models
- Ensure strong elicitation and quality control
- Analyze evaluation results
- Develop and implement safety mitigations, products, and tools
Benefits
- Health insurance coverage
- Retirement plans adapted to the location
- 50% transportation-home-to-work coverage
- Productivity expenditures up to €2k annually
- Office space if relevant
Hiring Process
Initial screening → first interview → paid work test → interviews → final 3-day paid work trial