Enterprise Experience

Scaled forecasting and operational intelligence across a fast-changing fulfillment network.

• ML platform owner on Databricks / Azure ML Foundry across forecasting, ETA prediction (quantile regression, conformal calibration), classification, outlier detection, and NLP.
• Built ETA prediction (100M+ orders/year) and demand forecasting (1M+ timeseries, weekly) end-to-end.
• Layered in anomaly detection, promotion adjustments, and stock-out corrections to keep forecasts honest under real operating noise.
• Mentored scientists and engineers; translated technical work for product and operations stakeholders.

Owned real-time pricing systems where quality and uptime had direct financial consequences.

• Owner of the ML training and reporting pipelines powering BIN, Market Cost, Market Rate
• Led automated tender acceptance and a baseline market-cost service that matched ensemble accuracy on standard orders using only five shipment inputs.
• Built REST services on AKS with batch endpoints and explainability features for human review.
• Interviewed, hired, and mentored as the team grew.

Delivered pricing and recommendation models that were useful in production, not just accurate offline.

• Deployed pricing models using gradient boosting, quantile regression, and ensembles.
• Tuned recommendation systems for freight matching and demonstrated measurable lift in production.
• Shipped explainability endpoints that translated model behavior into human-readable reasons.

Started in forecasting and experimentation while learning how data products succeed inside real organizations.

• Built enterprise forecasts using ARIMA and ETS across business accounts.
• Automated freight matching workflows in R and Python alongside DBAs, scientists, and engineers.
• Designed and ran A/B tests for recommendation approaches in a live matching application.

Applied predictive modeling in a noisy domain where positive outcomes were rare and precision mattered.

• Led model development across debt portfolios in healthcare, utilities, and other domains.
• Built classification and regression models to estimate payment likelihood and expected value in highly imbalanced data.
• Deployed a scoring workflow for incoming placements and tuned collection strategy from the results.