Leveraging machine learning to collect less sensor data for lightweight analytics pipeline

The Pennsylvania State University, Lightscline, Pennsylvania, United States

Keywords: Deep learning, machine learning, sensor, image, video, compressed sensing

This poster will present novel AI-based techniques for analyzing large volumes of sensor data by collecting less data upfront while preserving analytics accuracies. Collecting 10x less data upfront significantly reduces the data infrastructure and human capital costs associated with large scale data analytics and machine learning. The poster will walk through Lightscline's 250+ customer discovery interviews using the National Science Foundation I-Corps National Teams award and communicate the value proposition of this technology to stakeholders in the defense community. 2 specific value props are: 1. The VP of Technology of an oil & gas company wants to use our data-reduction AI to reduce their annual cloud processing costs for 4 machine learning from $250k to $25k. 2. The Analytics Manager of a Fortune 150 company wants to use our data-reduction AI to reduce their cloud costs for training ML models and make their edge software faster.