Researcher, Alignment Training
Core
Researching how training choices shape aligned behavior in frontier models by defining behaviors, designing data and training interventions, and building evaluation loops.
Role type
Senior researcher (alignment training)
Builds
Synthetic data methods, training interventions, evaluation loops, and data generation/filtering pipelines.
Domain
AI safety, large-scale model training, alignment
Deliverable
production ML models
Required skills
large-scale ML, pre-training, post-training, synthetic data, model evaluation, training infrastructure, experimental design, hypothesis formulation, pipeline building, result analysis
Preferred skills
judgment on research questions, evidence-driven work, practical execution
Technologies
large-scale ML frameworks, training pipelines, evaluation systems
Responsibilities
Develop synthetic data methods for higher-level behavioral tendencies; study pre/mid/post-training impacts on model behavior; build evaluation loops connecting behavior to objectives; design reusable data generation and filtering pipelines; create experiments distinguishing durable behavior from artifacts; collaborate across teams to translate insights into model behavior.
Seniority
Senior, hands-on IC