My role
Data scientist — causal inference and statistical methods.
Overview
Designed the analysis layer for the company experimentation platform. Implemented variance-reduction (CUPED), sequential testing to allow safe early stopping, and heterogeneous treatment-effect estimation to find which segments actually respond. Wrote guardrail-metric checks that halt a rollout automatically, and paired randomised tests with quasi-experimental methods for launches that could not be cleanly randomised.
Role & impact
Impact
Standardised 120+ experiments a quarter and caught three revenue-negative launches before full rollout.