Forward Deployed Research Scientist

Responsibilities

• Engage directly with frontier lab research teams in scoping meetings, challenge assumptions, and shape project specifications based on data composition effects on model outcomes.
• Develop deep scientific understanding of client architectures, training methodologies, and target capabilities to reason about data strategies, identify risks, and iterate empirically.
• Run ablation studies and fine-tune open-weight models on client data to validate data impact on model performance.
• Consult on workflow and quality systems, reviewing annotation schemas, task designs, and quality rubrics with Human Data Operations.
• Collaborate with Applied Research on publications, benchmarks, white papers, and conference submissions using client-grounded findings.

Requirements

Required:

• MS or PhD in Machine Learning, NLP, Computer Science, or related quantitative field.
• Hands-on experience fine-tuning large language models (e.g., Llama, Mistral, Qwen).
• Strong understanding of LLM training pipelines (pretraining, supervised fine-tuning, RLHF/DPO) and data quality/composition effects.
• Experience designing/executing rigorous experiments (hypothesis formation, controlled comparisons, statistical analysis).
• Ability to operate at speed (problem to results in days).
• Strong written/verbal communication for client presentations and publications.

Strongly Preferred:

• Experience at frontier AI lab, applied ML startup, or research with client interaction.
• LLM evaluation/benchmarking (metrics, eval harnesses).
• Human data pipelines (annotation workflows, quality assurance, inter-annotator agreement).
• Reinforcement learning, reward modeling, or RLHF.
• Published ML/NLP research.

What Matters:

• Applied instinct, comfort with ambiguity, cross-functional fluency, intellectual honesty.