All projects
Client project for Sanmina
Demand Forecasting System
An API-driven forecasting service that produces item-level weekly forecasts from BigQuery or uploaded data and turns them into client-ready Excel reports.
Forecasting & tabular ML
Overview
An end-to-end, API-driven pipeline for Sanmina that forecasts demand item by item, either from BigQuery enterprise data or from datasets a client uploads. Forecasts are reconciled with quarterly plans and delivered as structured Excel reports with capacity, inventory and supply context.
The problem
Planning needed item-level forecasts that still agree with quarterly business demand plans, drawn from many sources (demand, MDS, capacity, WIP, inventory, CTB and calendar data), without anyone moving data around by hand.
What I did
- Built the pipeline end to end and designed its architecture, with ingestion, preprocessing, forecasting, reconciliation and reporting as separate modules.
- Developed the modular FastAPI service supporting both automated cloud inference and file-based forecasting.
- Built the automated report generation engine.
How it works
- Preprocessing that turns demand, MDS, capacity, WIP, inventory, CTB and calendar data into clean, continuous ITEM × WEEK time series.
- MLForecast (XGBoost) models for large-scale item-level forecasting, with configurable start dates and dynamic horizons.
- Hierarchical reconciliation aligning weekly item forecasts with quarterly demand plans.
- Secure BigQuery integration for fully automated enterprise inference.
- Excel reports that combine forecasts with capacity, inventory and supply-pipeline insights.
Outcome
- Fully automated inference from BigQuery, removing the need for manual data handling.
- Client-ready reports produced directly by the pipeline.