Lead Data Scientist - Propensity & Segmentation (Telecom)Location: 100% Remote or Onsite in Dallas, TX MoI: Video for Remote OR F2F for Onsite for Dallas, TX Duration: 12+ monthsExperience: 15+ years (MUST HAVE)MUST HAVE Skill: (NON_NEGOTIABLE)Telecom domain mustData warehousing SQL Spark framework, pysparkComplexBig data conceptData science conceptCore DS Fundamentals – ml datascineceAdvanced Cloud SQL & TuningBusiness-Centric EvaluationPython EcosystemREQUIRED MACHINE LEARNING & EXPERIENCEExperience: 15+ years of professional experience as an applied Data Scientist building and deploying supervised and unsupervised machine learning models.Core DS Fundamentals: Deep understanding of traditional ML theory, including class imbalance mitigation, feature selection, probability calibration, and experimental design.Business-Centric Evaluation: Ability to evaluate models beyond standard AUC/ROC, focusing on lift charts, precision-recall curves, tier separation, and financial ROI.Python Ecosystem: Advanced proficiency in Python, specifically utilizing the traditional data science stack (pandas, NumPy, scikit-learn, XGBoost, LightGBM) within notebook and script-based workflows.TELECOM & GEOSPATIAL REQUIREMENTS (MUST HAVE)Telecom Domain Expertise: 3+ years specifically navigating telecom, broadband, wireless, or subscription-based data structures (e.g., understanding ARPU, churn cycles).Geospatial Literacy: Practical experience using spatial SQL functions (e.g., BigQuery GIS, PostGIS, H3/S2 spatial indexing) to join and analyze location-based data like lat/long coordinates, wire centers, or census tracts.WHAT YOU WILL DOHands-on Feature Engineering: Write, debug, and optimize complex SQL queries on cloud data warehouses. You will build clean feature sets from raw, massive source tables spanning customer billing, network performance, competitive footprint, and geographic data.Predictive & Behavioral Modeling: Build, calibrate, and maintain propensity and "take rate" models utilizing gradient boosted trees (e.g., XGBoost, LightGBM) to optimize marketing spend.Customer Archetypes: Develop unsupervised clustering and segmentation frameworks to group customers and addresses, enabling hyper-personalized marketing workflows.Enforce Core DS Rigor: Engineer features utilizing strict time-series windows to rigorously protect against data leakage, lookahead bias, and overfitting.Model Explainability & Performance: Evaluate and explain model mechanics using SHAP and feature importance. Monitor models in production to detect and remediate data and concept drift.Experimental Design: Collaborate with marketing teams to design A/B tests and randomized control trials (RCTs) to measure true incremental lift and isolate campaign performance from organic consumer behavior.Deliver Actionable Outcomes: Cleanly package outputs into business-ready deliverables, including feature dictionaries, performance tier charts, and scored target lists.