Member of Technical Staff - Pretraining

📋 Description

• Lead large-scale pretraining for multimodal models (image/video/audio).
• Shape training objectives, architectures, data strategies, and systems.
• Contribute across low-level GPU, systems, and high-level design.
• Lead focused research projects and cross-team initiatives.

🎯 Requirements

• Led or co-owned pretraining for a foundation model in production/release.
• Own architectural calls: attention, modulation, loss, tokenization.
• Deep experience with distributed training on 500+ GPUs.
• Strong intuition for architecture and objective design.
• Record of shipping: top venues + production impact.
• Deep Python and PyTorch; reading/modifying training code.

🎁 Benefits

• Note: No salary/compensation details included in this section.