We're hiring a Python‐first, statistics‐strong Data Science Intern (or junior ML Engineer) to partner with our Data Analysts on the hardest problems we have: probabilistic LTV, churn prediction, and forecasting that actually moves CAC, payback, and ROAS.
Envíe su solicitud a continuación después de leer todos los detalles y la información de apoyo sobre esta oportunidad de trabajo.
READ THIS FIRST — Apply only if you meet this: Eligibility: You're currently studying at a university in Spain (ideally in or near Barcelona) and can sign a "convenio de prácticas".
This position is open exclusively for internship agreements.
If you're not sure whether your university can provide one, please confirm with them before applying, as we'll automatically filter out candidates who can't meet this requirement.Otherwise, feel free to check out our other junior roles on the Carrots Lab LinkedIn page › Jobswhere a "convenio" isn't required.
WHAT YOU'LL OWN (Your Mission)LTV Prediction & Probabilistic ForecastsBuild, compare, and iterate LTV models (e.g., survival/retention‐based, zero‐inflated + Gamma, GBMs).
Produce calibrated predictions we can trust for UA and pricing.Churn & Retention ModelingPredict churn windows and retention curves by app/geo/cohort to inform payback and creative/keyword strategy.Signal Engineering for UATranslate model outputs into conversion signals (e.g., value‐based bidding targets) that improve Google/Meta optimization.Revenue & Cohort Analytics (Python)Pull, transform, and join event & purchase data (RevenueCat, AppsFlyer, Firebase, BigQuery) into model‐ready datasets.Decision ScienceEvaluate A/B tests (frequentist or Bayesian), pricing experiments, and funnel changes with confidence intervals and practical uplift.
WHAT SUCCESS LOOKS LIKE (90 Days)LTV v1 shipped per flagship app with baseline vs.
model comparison (e.g., MAE/RMSE or calibration plots) and a documented retrain cadence.Churn or retention model v1 that beats naive benchmarks (e.g., 30‐day retention mean) and informs UA payback assumptions.Signal spec delivered: which in‐app events/values we should pass to ad platforms, how to weight them, expected CPA impact.Reproducible codebase: clean notebooks/scripts with README, functions, and basic tests; data pulls automated in Python. xqysrnh Live dashboards (with Sergi): LTV/payback & cohort views wired to Looker for weekly decision‐making.