Testimonials
As an optimizer, Perforated AI doesn’t build networks, we make yours better. This page is dedicated to the experiences of our users and what Perforated AI has been able to do for them.
Our latest customer story comes from a user who improved the predictive accuracy of the AI model by 39%. The Average Root Mean Squared Error (difference between model’s prediction of required supply and the actual requirement) decreased from 27.7 to 16.9 after one week of experimentation with PAI.
Andres Enrique Marín Zambrano is a graduate student at University of Texas At San Antonio, working on AI applications for Supply Chain Prediction.
“It’s a bad scenario in supply chains to have unused excess inventory or, even worse, no inventory at all. This leads to lost work hours in manufacturing and missed sales in retail, resulting in customer dissatisfaction,” he told us.
Andres was leveraging an AI agent named ClaudIA to forecast demand in order to optimize working supply, ensuring better inventory availability. ClaudIA utilizes a strategy that incorporates Market Basket Analysis and RFM (Recency, Frequency, Monetary) Analysis, combined with a neural network architecture that includes Long Short-Term Memory (LSTM) and a Multilayer Perceptron (MLP) to generate multi-item forecasts.
““In just a week we were able to improve ClaudIA. With Perforated AI now ClaudIA is giving better results achieving outstanding forecasting for demand planners in the supply chain industry.””
Perforated AI’s CEO, chose to be hands-on with our first user. He worked with Andres along the way to get Perforated Backpropagation™ integrated into ClaudIA while taking his feedback to update our README’s and API. Andres said regarding the integration process, “After reading the documentation and with some guidance everything went smoothly.”
After the first run showed promising results, Andres worked with Perforated AI’s system to further improve his model. After a week of experimentation Andres was able to reduce the Average Root Mean Squared Error across the 18 items in the dataset from 27.7 to 16.9, a whopping 39% improvement.
