Built and led full product lifecycle of marketing data science solutions for portfolio of clients including JP Morgan, Cisco, and Republic Services.• Media performance and product demand time-series forecasting using classical statistics (Holt-Winters Exponential Smoothing, ARIMA, Moving-Average, Autoregressive) and Deep Learning (Neural Prophet, Dense NN, Convolution N, LSTMs, and Auto-Regressive LSTMs).• Replaced GLM MMMs with ensemble models (Random Forest and XGBoost) to capture non-linear revenue attribution with SHAP Values (XAi) and fairly attribute incremental gains to marketing channels. Produced "importance" impact on revenue, diminishing returns curves, minimum investment/ positive impact thresholds, and channel interaction effects. Produced 12.8% increase in ROAS.•In strategic partnership with Google, Led Value-based bidding (VBB) GCP based solution which maximized ad conversion value at auction by incorporating both online and offline conversion values, as well as revenue and non-revenue KPIs, to forecast the probabilistic lifetime value of conversion for current and future customers. Deployed Bayesian BG-NBD and DNN CLV models within a TFX Kubeflow pipeline using a fully managed Kubernetes cluster in GCP.• Owned client analytics within several development work streams from KPI definition/designing measurement frameworks to automation of ongoing reporting and insights.

Led and performed the following media measurement solutions:•Elevate the work of the whole team by automating need to know KPIs using SQL and Python enabling deeper insight behind the why's of KPI movement.• Paid Media optimization lifts using time series intervention causal inference with bayesian structural time series model (BSTS).• Built Dash web application to provide an automated media taxonomy quality assurance solution that managed the design and implementation of taxonomies. Improved dataflow efficiency by 95% and accuracy by 98%. • Using Snowflake, Big query, and Redshift databases, built Datorama, Tableau, and Power BI dashboards to support ad-hoc stakeholder questions.

• Designed and implemented a multi-class logistic regression classification for social sentiment sensitivity using frequency based NLP data mining methods to support social brand reputation management• Run analysis to support ad-hoc inquiries by social media SME's