Back to Projects

Predictive Demand Engine.

AI-driven forecasting tool for supply chain optimization, utilizing LSTM models to predict inventory needs with 94% accuracy.

Python
TensorFlow
PostgreSQL
FastAPI

Overview

This engine processes TBs of historical sales data to forecast future inventory requirements. By using Long Short-Term Memory (LSTM) networks, it captures seasonal trends and long-term dependencies that traditional linear regression models miss, directly integrating with the ERP system.

Technical Challenge

Training on heterogeneous time-series data with varying sequence lengths. I implemented dynamic padding and masking layers within the TensorFlow pipeline, allowing the model to efficiently batch and train on disparate product lifecycles without skewing the weights.