Artificial Intelligence / Machine Learning

Data Engineering

Lyteworx specializes in automating the Extraction, Transforming, and Loading process for massive amounts of data for use in supervised learning. We develop code to interact with custom APIs across multiple applications to keep the data flowing prior to and after labeling.

Model Development

Lyteworx has produced computer vision Convolutional Neural Network models for object detection on high altitude imagery.  The model was trained using the ResNet-50 CNN architecture starting from a pre-trained model using the Microsoft COCO dataset and finished the training using the Cars Overhead With Context (COWC) dataset, provided by the Lawrence Livermore National Laboratory.

We take industry best-practices to include openly available, high-performance models and incorporate their functionality into customer specific applications.  This process can include retraining the model (transfer learning from existing pre-trained model) with custom data.