Portfolio Project by Wil Low / Draculess99

Autonomous & Multi-Agent AI Warehouse Workforce Forecasting

This project started on the warehouse floor. After two years as a fulfillment center dispatcher watching staffing decisions happen reactively — overtime called too late, VTO offered too early — I built a system to get ahead of it.

An advanced workforce planning system that forecasts weekly warehouse workload, recommends VET, VTO, or Normal staffing actions, evaluates risk and cost impact, and explains decisions through agentic AI and an autonomous supervisor workflow.

Wil Low / Draculess99 — VET/VTO Warehouse Forecasting, Workforce Analytics, Demand Forecasting, Agentic AI, Autonomous Supervision, and AI-Powered Labor Optimization.

Project Overview

This project upgrades a traditional VET/VTO workforce forecasting dashboard into a multi-agent decision intelligence system. Instead of only predicting weekly workload, the application routes forecast results through specialized AI workflow components for staffing recommendations, cost impact analysis, operational risk review, and executive decision summaries. The result is a warehouse labor planning system that can explain why VET, VTO, or Normal staffing is recommended — and connect those recommendations to business impact, risk, and operational readiness.

Problem Forecast weekly warehouse workload and identify staffing pressure before it becomes reactive.
Decision Output Recommend VET, VTO, or Normal staffing based on forecasted demand, risk, and cost impact.
Business Goal Support labor planning, reduce reactive overtime, control cost, and improve workforce readiness.

Project Highlights

Business Value

Warehouse labor planning is often reactive: managers respond to workload spikes, overtime pressure, and staffing gaps after they have already affected operations. This project demonstrates how forecasting, agentic AI, and autonomous workflow supervision can support earlier labor decisions by connecting demand prediction with staffing actions, cost impact, operational risk, and executive-level explanation. Instead of producing only a forecast, the system converts workload signals into VET, VTO, or Normal staffing recommendations that are easier for operations teams to review, explain, and act on.

Business Impact / Demonstration Run

The system forecasts labor demand and generates VET, VTO, or Normal staffing recommendations. A cost-impact model then estimates the potential labor cost implications of each staffing decision compared with a baseline staffing strategy.

Forecast Horizon Up to 30 weeks, depending on the selected scenario.
Decision Output VET, VTO, or Normal staffing recommendation.
Estimated Cost Impact Model-generated estimate from the selected forecast period.

Designed for live deployment; demonstration runs simulate warehouse conditions based on two years of dispatch operations experience.

Model Performance and Estimated Cost Impact

Notebook backtesting showed that the XGBoost recursive forecasting model reduced forecasting error by approximately 53% compared with a seasonal naive baseline.

Seasonal Naive RMSE 2,006,104
Lasso RMSE 1,385,610
XGBoost RMSE 941,679
Forecast Error Reduction 53.1% vs Seasonal Naive
Estimated Cost Reduction $546,855
Estimated Percentage Reduction 78.2%

These figures are model-generated estimates from notebook backtesting and demonstration scenario runs. They are not measured savings from a live warehouse deployment.

Technical Architecture

The system combines machine learning forecasting, backend API services, dashboard visualization, and an autonomous multi-agent decision workflow. Forecast results are passed through specialized workflow nodes for staffing recommendation, cost impact analysis, risk review, memory/context handling, and executive summary generation.

VET/VTO workforce forecasting system architecture diagram

Architecture diagram showing the Streamlit dashboard, Flask forecasting API, XGBoost model, staffing recommendation logic, cost impact layer, and agentic workflow components.

Technology Stack

Python XGBoost Machine Learning Time-Series Forecasting Demand Forecasting Workforce Analytics VET/VTO Optimization Streamlit Flask API Docker AWS ECS Fargate Railway Render LangGraph Workflow Agentic AI Autonomous Supervisor Multi-Agent System Risk Monitoring Cost Impact Analysis Executive Decision Summaries Predictive Analytics Warehouse Operations

Links

Open the Advanced VET/VTO Multi-Agent App

```