Senior Data Scientist: Predictive Operations & Experimentation

O projekcie: About the Role We are looking for a Senior Data Scientist to join the Nextbike City Insights (NCI) team, where we build the operational tooling suite that powers how cities run their bike-sharing networks. This person will sit at the intersection of predictive modelling and rigorous data analysis, working directly with operational data to build models that improve city operations, and then design the experiments that prove those models work. The ideal candidate is equally comfortable training a gradient-boosted demand forecaster as they are designing an A/B test to validate its real-world impact on rebalancing costs. Key Responsibilities Predictive Modelling & Algorithms - Design, build, and deploy predictive models for core operational challenges, starting with demand forecasting and predictive rebalancing at the station level. - Integrate heterogeneous signals (historical trip data, weather, events, seasonality, urban patterns) into production-grade feature pipelines. - Continuously evaluate and iterate on model performance using backtesting and live monitoring. - Explore and prototype adjacent model opportunities: maintenance prediction, fleet sizing optimisation, dynamic pricing sensitivity. Data Analysis & KPI Development - Propose and define KPIs and operational metrics that reflect real business impact (e.g., rebalancing cost per trip, station availability rate, idle time reduction). - Perform deep-dive correlation studies and exploratory analyses to surface actionable patterns in operational data. - Formulate and stress-test hypotheses before committing resources to live experimentation. - Build dashboards and automated reports that make model outputs and KPI movements legible to non-technical stakeholders. Experimentation & Causal Inference - Design statistically rigorous experimentation frameworks (A/B tests, switchback experiments, difference-in-differences) that can be deployed across live city operations. - Define sample sizes, control groups, and success criteria before experiments launch. - Analyse experiment results, quantify uplift with confidence intervals, and translate findings into clear go/no-go recommendations. - Own the feedback loop: model → hypothesis → experiment → measurement → model refinement. - Tackle the unique challenges of experimentation in networked, real-world systems: account for spatial spillover effects between nearby stations, design cluster-randomised or switchback experiments that minimise contamination, and handle interference where treating one unit (station, zone, city) affects outcomes at others. - Develop strategies for isolating treatment effects when supply and demand are shared across a network (e.g., rebalancing one station changes availability at neighbouring stations). - Apply design-of-experiments principles to constrained environments: limited number of cities/zones, high variance, non-independence of observations, and operational constraints on what can be randomised and when. What We Offer - Direct impact on how cities move. Your models run in production, not in notebooks. - Ownership of the full data science lifecycle: exploration → modelling → experimentation → deployment. - A small, senior team where your work shapes product and operational strategy. - Flexible remote setup within Poland-compatible time zones. Wymagania: Must-Have - 5+ years of hands-on experience building predictive/ML models on real-world operational or time-series data. - Strong proficiency in Python (pandas, scikit-learn, XGBoost/LightGBM) and SQL. Our data warehouse runs on Snowflake. - Demonstrated experience designing and analysing controlled experiments (A/B tests, quasi-experimental methods). - Solid grounding in statistics: hypothesis testing, regression, confidence intervals, power analysis. - Ability to translate ambiguous business problems into quantifiable hypotheses and measurable outcomes. - Clear communicator who can present model logic and experiment results to operations teams and leadership. Nice-to-Have - Experience with geospatial data, mobility/logistics, or marketplace-style supply-demand problems. - Familiarity with causal inference techniques (uplift modelling, instrumental variables, synthetic control). - Exposure to MLOps tooling (MLflow, Airflow, dbt, or similar) for model deployment and monitoring. - Experience working with weather APIs, event data, or other external signal sources. - Background in operations research or optimisation (e.g., vehicle routing, scheduling).