Past work · Machine learning

Machine learning lead-time model

I developed a model and Power BI experience that turn intercontinental shipment history into explainable arrival windows, risk alerts, and operational decisions.

The work

Prediction designed for day-to-day use.

The model analyzed 18,000 historical container shipments and improved prediction accuracy by 43%. The surrounding dashboard lets sales and supply teams search a container, understand the model’s confidence, spot at-risk arrivals, and examine the attributes influencing lead time.

Tracking Operations Output Lookup Tool dashboard screenshot — **Tracking Operations Output Lookup Tool**A quick-answer search experience where users select a container and choose multiple questions about it.
This dashboard uses simulated data to protect confidential company information.

Container List dashboard screenshot — **Container List**A bookmarkable operating list of monitored containers and their predicted arrival windows.
This dashboard uses simulated data to protect confidential company information.

Incoming Arrivals dashboard screenshot — **Incoming Arrivals**A sales-planning view showing which part numbers are expected to arrive week by week.
This dashboard uses simulated data to protect confidential company information.

Container Finder dashboard screenshot — **Container Finder**Shows a container’s status, current journey stage, predicted arrival window, and confidence.
This dashboard uses simulated data to protect confidential company information.

Container Details dashboard screenshot — **Container Details**Drill-through detail using similar shipments, nearest-neighbor logic, and model output to explain the arrival window.
This dashboard uses simulated data to protect confidential company information.

Alert Containers dashboard screenshot — **Alert Containers**Prioritizes containers that may miss their expected timing so teams can intervene earlier.
This dashboard uses simulated data to protect confidential company information.

Performance Over Time dashboard screenshot — **Performance Over Time**Compares ocean-carrier lead-time performance over time against the full shipment population.
This dashboard uses simulated data to protect confidential company information.

Attribute Impact dashboard screenshot — **Attribute Impact**Quantifies how a shipping attribute changes average lead time, making model recommendations easier to explain.
This dashboard uses simulated data to protect confidential company information.

Health Navigation dashboard screenshot — **Health Navigation**Routes users to the model-health views for performance, confidence calibration, and weekly hit rate.
This dashboard uses simulated data to protect confidential company information.

Model Stats dashboard screenshot — **Model Stats**Compares the ensemble model against operational baselines using hit rate, error, and bias.
This dashboard uses simulated data to protect confidential company information.

Confidence Levels dashboard screenshot — **Confidence Levels**Checks whether each confidence band corresponds to the frequency with which shipments actually arrive in window.
This dashboard uses simulated data to protect confidential company information.

Hit Rates dashboard screenshot — **Hit Rates**Tracks weekly performance for each model and baseline to make changes in model health visible.
This dashboard uses simulated data to protect confidential company information.

Glossary dashboard screenshot — **Glossary**Defines model terms so business users can interpret the dashboard consistently.
This dashboard uses simulated data to protect confidential company information.