2026 Kleiner Perkins Engineering Fellow
Kleiner Perkins runs a fellowship that places engineers inside its portfolio companies for a summer. Mine was Inkitt, a Series C storytelling startup whose apps, Galatea and CandyJar, serve millions of recommendations a day. I worked on the machine learning that decides what gets shown and what a reader is worth.
I architected a reusable pLTV revenue-prediction system: a two-stage hurdle model with a calibrated CatBoost head, separating whether a user will ever pay from how much they will pay. It reached 7.2x top-decile lift on a 2.2M-user cohort and now runs daily through an Airflow pipeline that feeds CRM and marketing.
I then rebuilt the core recommendation engine behind both apps, replacing standard collaborative filtering with a custom hybrid model. Click-through rose 18% and engagement 24% across millions of daily recommendations.
I also built and deployed the production AI assistants that help readers discover stories and TV shows in each app.