Research AI safety remotely from the United States with HFC and SLPreP, focusing on evaluations, monitoring, and agent behavior.

This part-time residency supports researchers pursuing or holding a PhD in computer science, machine learning, statistics, mathematics, or related fields. Residents investigate empirical foundations of AI safety, guided weekly by an SLPreP researcher while driving daily work. Projects may address AI control, alignment red-teaming, multi-agent safety, recursive self-improvement, evaluations, or misuse risks.

Key Responsibilities:

- Turn safety questions into hypotheses, baselines, and controlled experiments
- Implement and debug evaluation harnesses, environments, datasets, and monitoring methods
- Identify confounds and interpret negative research results




- Propose next experiments with high-level guidance
- Produce reproducible code, logs, analysis, and technical writing

Requirements:

- Pursuing or completed PhD, or equivalent research track record
- Interest in technical AI safety research
- Ability to progress independently with high-level guidance
- Explicit research communication skills
- Responsible handling of confidential information and internal resources

Build empirical AI safety research through evaluations, monitoring, controlled experiments, and precise technical communication.

📌 AI Safety Research Resident (Canada)
🏢 Inested
📍 Canada