Member of Technical Staff - ML Scientist, Japanese Multimodal
Core
Improve capabilities and behavior of Liquid Foundation Models for Japan and the global market through post-training strategies.
Role type
Senior IC ML Scientist (post-training)
Builds
High-quality checkpoints and reusable methods for language and multimodal models
Domain
AI / Machine Learning / Multimodal Models
Deliverable
production ML models
Required skills
post-training modern language or multimodal models, machine learning fundamentals, post-training and RL methods, open-source ML ecosystem engineering, rigorous experimental design, training and evaluation data curation, research-to-implementation translation, agent usage
Preferred skills
preference optimization, reinforcement learning for foundation models, multimodal model post-training (text, vision, audio), synthetic data pipelines, reward models, verifiers, model-based evaluations, Japanese reading proficiency
Technologies
open-source ML ecosystem
Responsibilities
Design and execute post-training strategies including supervised fine-tuning, preference optimization, reinforcement learning, and distillation; Build and curate high-quality training data using human, synthetic, and model-generated signals; Develop evaluations to expose capability and reliability gaps; Conduct systematic error analysis to improve data mixtures and training methods; Run controlled experiments and ablations; Develop scalable training and evaluation pipelines
Seniority
Senior, hands-on IC