Member of Technical Staff Multimodal Understanding

You will advance multimodal understanding and generation across image, video, audio, and text. You will build and optimize distributed training, inference, tokenization, and data systems; develop evaluations and research tooling; and enable real-time multimodal reasoning, tool use, and agentic behaviors.

Responsibilities

• Design, build, and optimize distributed systems for multimodal training, inference, data processing, and tokenization.
• Develop high-throughput pipelines for multimodal data acquisition, preprocessing, filtering, generation, and management.
• Advance multimodal understanding, generation, alignment, world modeling, and reasoning capabilities.
• Drive data curation, filtering, quality studies, and scalable pipelines.
• Create evaluation frameworks, benchmarks, reward models, and metrics.
• Innovate on algorithms, modeling approaches, co-design, and scaling paradigms.
• Build research tooling, interfaces, prototypes, demos, and full-stack applications.
• Enable reasoning, tool calling, agentic behaviors, orchestration, and real-time interactions.

Requirements

• Experience with multimodal pre-training, post-training, or fine-tuning.
• Expert proficiency in Python.
• Experience with JAX, PyTorch, or XLA.
• Experience building or optimizing large-scale distributed ML systems.
• Experience designing and running large-scale data pipelines.
• Knowledge of evaluation design, benchmarks, reward modeling, or reinforcement learning techniques.

Benefits

• Equity
• Medical coverage
• Vision coverage
• Dental coverage
• 401(k) retirement plan
• Short-term disability insurance
• Long-term disability insurance
• Life insurance
• Discounts and perks