The success of our customers within the Federal, Defense, and Intelligence communities is increasingly dependent on their ability to unlock the knowledge and insights contained in the ever-expanding pool of data.

We use many different processes and analytics, such as statistical analysis, data mining, and predictive analytics, to extract actionable information from structured and unstructured data in various forms stored across the enterprise. Our methods include training machine learning models, scaling inference pipelines, creating tools facilitating labeling campaigns, link analysis, multi-criteria decision analysis, and natural language processing.

Our teams aren’t just doing this in an R&D environment; we are implementing these algorithms at scale and deploying them for our analysts in the Federal, Defense, and Intelligence communities. We are using a wide range of data types to support our algorithms, to include: commercial and national imagery, Synthetic Aperture Radar (SAR), geospatial foundation data (e.g., buildings and roads), and moving object data, among others.

Our teams are tackling tough problems such as:

• Ensuring the data is where it needs to be, when it needs to be there.
• Scaling computer vision algorithms to work at the rate data is collected
• Mining databases of vehicle positional data to identify known or anomalous behaviors
• Creating and orchestrating pipelines for automated broad area object detection
• Automating the Quality Assurance (QA) of critical geospatial foundation data

Some of the open source and commercial software technologies we use include:

• Spark, Spark ML
• TensorFlow
• CUDA/CUDNN
• Caffe
• ONNX
• PyTorch
• Keras
• Scikit learn
• Jupyter