Member of Technical Staff (Pretraining Systems)

Magic as a Research Engineer on the Pretraining Systems team, where you will design and operate the distributed infrastructure for training long-context models at scale.

Responsibilities

• This role focuses on large-scale model training across massive GPU clusters, requiring a deep understanding of parallelism strategies, performance trade-offs, and experience in multi-node GPU environments.
• You will optimize communication patterns, improve checkpointing and fault tolerance, and collaborate with research teams to align model architecture with systems realities.
• Conception et exploitation de l'infrastructure distribuée pour l'entraînement de modèles à long contexte.
• Optimisation des schémas de communication et de la synchronisation des gradients pour améliorer les performances.
• Collaboration avec les équipes de recherche pour aligner l'architecture du modèle avec les réalités du système.
• Deep understanding of parallelism strategies and performance trade-offs

Requirements

• Track record of improving performance or reliability of large-scale systems
• Experience training large models in multi-node GPU environments
• Experience debugging cross-layer issues in production ML systems

Strong software engineering and distributed systems fundamentals

• Strong ownership mindset and ability to operate critical infrastructure
• We value quick learning and grit just as much as skill and experience

Pay

USD 275,000 - 550,000

Benefits

• Enjoy benefits such as unlimited paid time off, competitive salary with equity compensation, and comprehensive healthcare coverage.