At パーソルキャリア株式会社, I worked on recommendation systems and production machine learning in the recruitment industry. I replaced fixed recommendation thresholds with an optimization-based policy that kept recommendation volume stable while expanding coverage for users who had previously received fewer recommendations by approximately 20–25%.
In a month-long A/B test with tens of thousands of users per group, the treatment group had nearly 50% more successful hires. I also worked on KPI design, experimentation, ML experimentation infrastructure, and production monitoring using AWS, MLflow, and Evidently. In graduate school, I used hierarchical Bayesian models with RStan to predict pressure injury progression.
