Research, Safety
Core
Researching and implementing AI safety measures to ensure models are safe, trustworthy, and handle harmful or dual-use requests.
Role type
AI safety researcher (hands-on IC)
Builds
Safety-focused AI models, evaluation pipelines, and red-teaming mitigations
Domain
Artificial Intelligence / Machine Learning
Deliverable
production ML models
Required skills
AI safety research, RLHF/RLAIF, alignment and preference modeling, safety evaluations, red-teaming, Python, deep learning frameworks (PyTorch, TensorFlow, JAX), distributed training debugging
Preferred skills
Long-horizon/agentic task evaluation, synthetic data generation, modern red-teaming/jailbreaking techniques, AI safety publications, scalable oversight knowledge
Technologies
PyTorch, TensorFlow, JAX
Responsibilities
Build data filtering pipelines and quality classifiers for pre-training corpora; Apply post-training techniques (RLHF/RLAIF) to shape model behavior; Design and maintain safety evaluations for long-horizon and agentic tasks; Generate and curate synthetic data for safety training; Red-team models to surface failure modes and design mitigations
Seniority
Mid-to-Senior, hands-on IC