Senior Data Scientist, AI Product Insights

About the role

As the first Data Scientist embedded in product engineering, you'll champion integrating cutting-edge data science techniques into Mixpanel's products and serve as a methodological resource for cross-functional teams tackling problems that benefit from deeper DS expertise — such as adaptive experimentation. You won't just advise on Proactive Insights; you'll be the analytical brain driving how Signals, Forecasting, Simulation, Predictions, and Cohort Detection actually work. Your models are the reasoning layer behind an AI system that proactively tells customers what changed, why, and what to do next — and increasingly, the layer behind an agent that acts on their behalf. As more of this experience becomes agentic, rigorous causal grounding is what separates a trustworthy recommendation from a plausible-sounding one. You'll be the person who makes sure it's the former. You'll design and validate causal inference approaches that go beyond surface-level correlation, and partner on how those outputs get translated — often via LLMs — into clear, natural-language, actionable experiences for Mixpanel's customers: you own the rigor, the system owns the explanation. You'll partner closely with strong product engineers who own the implementation — your job is to make sure the methodology is rigorous, well-documented, and grounded in real outcomes. You'll also collaborate cross-functionally with teams like AI platform, analysis, and data infrastructure to scale your analytic solutions beyond what you could build alone. This is a high-impact, high-autonomy role on a small, fast-moving team. You'll have significant influence over the analytical direction of a new product category at Mixpanel that helps thousands of companies understand what truly drives their most important metrics.

Responsibilities

• Own the end-to-end analytical design for Signals, Forecasting, Simulation, and Cohort Detection — including methodology selection, statistical validation, and iteration based on results
• Assess data quality and trust prerequisites before extending forecasting or predictive features to customers — a model is only as trustworthy as the data feeding it
• Design and apply causal inference methods to move beyond correlation and establish which user behaviors genuinely drive downstream business outcomes
• Build and own time-series forecasting models that project KPI trajectories against goals — extending our existing use of TimesFM into customer-facing forecasting features
• Build survival analysis and retention models that underpin Signals and Simulation outputs
• Develop clustering and behavioral similarity approaches for Cohort Detection that are both statistically sound and interpretable to end users
• Document methodology clearly — including assumptions, validation approaches,