AnthropicSan Francisco, CAConducts mechanistic interpretability research to reverse-engineer language models, developing methods to understand neural network algorithms for AI safety. Requires scientific research background, Python proficiency, and collaborative engineering mindset.
Familiarity with Python is required.
Education requirements: At least a Bachelor's degree in a related field or equivalent experience.
Conducts research on pre-training methodologies for large AI models, develops new architectures and data strategies, runs large-scale experiments, and publishes findings. Requires strong ML fundamentals, Python proficiency, and experience with deep learning frameworks.
Conducts research to advance audio capabilities in AI models, designing and training large-scale multimodal systems, building audio data pipelines, and publishing findings. Requires ML expertise, Python proficiency, and experience with deep learning frameworks.
Conducts experimental ML research on AI alignment and safety for powerful systems, focusing on scalable oversight, control, and stress-testing. Requires strong software/ML engineering, empirical research experience, and Python proficiency.
Research Engineer implements and scales post-training techniques like Constitutional AI and RLHF for production AI models, optimizing capabilities, alignment, and safety. Requires strong Python skills, ML systems experience, and ability to handle complex distributed training pipelines.
Research Engineer optimizes and scales production pretraining of frontier AI models, handling performance, debugging, experiments, and on-call incidents. Requires expertise in JAX, TPU, PyTorch, or large-scale ML systems with a 50/50 research-engineering balance.