Research, Pre-Training Science
Core
Advancing the science of how large models learn from data by exploring new pre-training methods, architectures, and learning objectives.
Role type
Pre-training Research Scientist
Builds
Large-scale AI models and foundational research
Domain
Artificial Intelligence / Machine Learning
Deliverable
production ML models
Required skills
Experimental design and analysis, distributed computing, Python, deep learning frameworks (PyTorch/TensorFlow/JAX), theoretical grounding in CS/ML/Physics/Math
Preferred skills
Probability and statistics, large-scale model training experience, publication record in pre-training, curriculum learning expertise, evaluation framework design
Technologies
PyTorch, TensorFlow, JAX
Responsibilities
Research and develop new methodologies for pre-training, design data curricula and sampling strategies, collaborate with infrastructure teams for large-scale experiments, publish and present research findings
Seniority
Individual Contributor (IC), research-focused