Safety Post-training
Core
Building shared systems and methods for safety post-training of frontier text, coding, and agentic models to improve model behavior at scale.
Role type
Member of Technical Staff (AI Safety Engineer)
Builds
Reproducible workflows connecting data generation, evaluation, training experiments, and production feedback.
Domain
AI Safety, Large Language Models, Responsible AI
Deliverable
production ML models
Required skills
Python, LLM training/fine-tuning, rigorous model evaluation, experimental design, data pipeline engineering, annotation design, automated graders
Preferred skills
Safety post-training methods (SFT, RLHF, RLAIF), RL infrastructure, adversarial behavior mitigation, cybersecurity, coding agent evaluation, statistical methods, human-evaluation design, model lifecycle platforms
Technologies
Python, distributed compute, cloud infrastructure, workflow orchestration