LMArena is looking for a Senior Data Scientist to explore and reason about the data that powers millions of AI evaluations each week, uncover insights into how frontier models behave in the real world, and improve the reliability and interpretability of their AI evaluation systems.
Requirements
- Strong proficiency in Python, with deep experience in Pandas, NumPy, and distributed frameworks like Spark.
- Expertise in statistical modeling, causal inference, and experimental design.
- Experience reasoning about data distributions, sample quality, and the effects of data distribution shifts.
- Background in AI model evaluation.
- Experience working with LLM outputs (for example, LLM-as-a-judge), embeddings, or other large-scale model artifacts.
- Experience with A/B testing.
Responsibilities
- Explore and analyze large, complex datasets to uncover patterns, biases, and causal relationships in model behavior and system performance.
- Formulate hypotheses about data quality, evaluation outcomes, and model performance — then design experiments to validate or refute them.
- Build reproducible analysis pipelines using Python, Pandas, NumPy, and Spark to process and interrogate large-scale data.
- Partner with ML researchers and engineers to design metrics and analyses that evaluate how models perform across domains, prompts, and tasks.
- Develop causal reasoning frameworks and statistical methods that help explain why models behave as they do — not just how well they perform.
- Communicate insights (for example, via blog posts) clearly to technical and non-technical partners, informing both research direction and infrastructure improvements.
Other
- 6+ years of experience in data science, ML analytics, or applied research, preferably in AI, ML, or large-scale data environments.
- Strong communication skills and the ability to collaborate closely with ML researchers and engineers.
- The opportunity to work on cutting-edge AI with a small, mission-driven team
- A culture that values transparency, trust, and community impact
