• Author complex, high-fidelity reasoning traces that document how an LLM should plan, use tools, and make decisions across sophisticated technical tasks
• Design structured traces that capture multi-step problem decomposition, logical inference, and real-world decision-making strategies
• Review and refine reasoning traces to ensure they reflect sound ML thinking and produce reliable, generalizable model behaviour
• Develop data strategies that help LLMs navigate intricate, open-ended scenarios with greater consistency and accuracy
• Contribute senior-level insights that elevate the overall quality and depth of training data

Who You Are

• Experienced in machine learning, AI research, or a closely related technical field — you understand how models learn and where they go wrong
• Skilled at breaking down complex problems into clear, logical, documented steps — the kind of thinking that makes great training data
• Familiar with LLM evaluation, fine-tuning, or training methodologies
• A precise, structured communicator who can make sophisticated reasoning legible and reproducible
• Self-directed and comfortable producing high-quality work independently

Nice to Have

• Prior experience with data annotation, evaluation systems, or AI quality assurance
• Top-tier Kaggle competition results (Grandmaster or Master level) — a strong signal of your model intuition and problem-solving depth
• Hands-on experience working with or evaluating large language models directly
• Background in AI safety, interpretability, or model alignment research

Why Join Us

• Work at the frontier — collaborate with world-leading AI research labs on projects that are shaping the future of machine intelligence
• Fully remote and flexible — work when and where you do your best thinking
• Freelance autonomy with the substance of genuinely meaningful, high-impact work
• Go beyond typical ML contracting — your expertise directly influences how frontier models reason and behave
• Potential for ongoing work and contract extension as new projects launch