Member of Technical Staff ML Scientist Japanese Multimodal

You will own the post-training experimental loop for language and multimodal models. You will develop training strategies, curate data, build evaluations, run controlled experiments and ablations, and analyze errors. You will create scalable training and evaluation pipelines and turn successful methods, datasets, and tooling into measurable model improvements.

Responsibilities

• Design and execute post-training strategies for language and multimodal models
• Build and curate training data using human, synthetic, and model-generated signals
• Develop evaluations for capability and reliability gaps
• Conduct error analysis to improve data, objectives, training methods, and model behavior
• Run controlled experiments and ablations
• Develop scalable training and evaluation pipelines
• Contribute post-training methods, tooling, datasets, and findings

Requirements

• Experience post-training modern language or multimodal models
• Knowledge of machine learning fundamentals, post-training, and reinforcement learning methods
• Engineering skills and proficiency with the open-source ML ecosystem
• Experience running rigorous experiments, baselines, ablations, and error analysis
• Experience building, curating, or assessing training and evaluation data
• Ability to create reliable implementations and measurable model improvements
• English proficiency
• Experience using agents to amplify work

Benefits

• Flexible working hours
• Unlimited paid time off
• Equity
• Standard benefits for employees in Japan