AI/Machine Learning Expertise & Capabilities

ASRC Federal’s machine learning, advanced analytics and data science suite of tools provides relevant and actionable intelligence to decision-makers — solving complex business and mission challenges.

Overview

Our unique solutions consist of customer-driven implementations. We work to understand your unique challenge in order to tailor-fit a solution to meet your needs. All of our solutions are based on industry best practices and leverage open source frameworks. Our team has extensive domain knowledge with advanced machine learning techniques and expertise to ensure our clients achieve mission success.

•  Machine Learning for Space Missions
•  Machine Learning for Business Systems
•  Machine Learning at the Tactical Edge

Download Overview
Schedule a Demo    

Key Applications

  • Clustering Icon
    Clustering Optimization
    • Unsupervised learning library used to draw inferences
    Clustering Optimization
    • Unsupervised learning library used to draw inferences
    Clustering Icon

    Unsupervised learning library used to draw inferences from datasets consisting of input data that has not been human labeled.

    Overview

    • Toolkit of capabilities for unsupervised learning used to draw inferences from datasets consisting of input data that has not been human labeled.
    • Divides data into meaningful and useful groups (clusters). These clusters can then be compared against each other, and outliers identified.
    • Membership History support tool observes clustered groups over repeated runs to identify stable memberships over time.
    • Includes optimizer for automated hyperparameter tuning and model selection using grid search, subsurface partition, or genetic algorithm.
    • Supports data visualization and plotting locally or headless via PyPlot.
    • Implemented as a Python library, leveraging scikit-learn and supports six forms of Manifold Learning and density based clustering (DBScan).
    • Configured with Gradle/Docker to run at scale in cloud environments.
    • Data connectors for Hadoop Distributed File System (HDFS), Amazon Simple Storage Service (S3).

    Mission Application

    • Deployed in production to identify anomalous behavior identified through medical claims records (potential indicators of fraud, waste, and abuse) within commercial health provider networks.
    • Used to identify similarities and identify latent anomalies in multi-dimensional complex data without well-characterized tagged data sets. These data sets can include human behaviors, enterprise financial data, employee time records, network traffic, to hardware monitoring to identify anomalous behavior in subsystems, or spacecraft within a mission or fleet. 

     

  • Data Analytics Workbench Icon
    Data Analytics Workbench
    • Data science tools used to develop/test machine models
    Data Analytics Workbench
    • Data science tools used to develop/test machine models
    Data Analytics Workbench Icon

    Overview

    • Powerful data science tools used to develop/test machine models and interact with static and real-time data streams directly on the deployed cluster.

    Mission Application

    • Easily develop, deploy and test machine models with an interactive data analytics workbench with a web-based notebook and machine learning engine.

     

     

  • Group Polarization Agent-Based Simulation Icon
    Group Polarization Simulation
    • N-dimensional scalable simulations for clustering
    Group Polarization Simulation
    • N-dimensional scalable simulations for clustering
    Group Polarization Agent-Based Simulation Icon

    N-dimensional scalable simulations for clustering benchmarks, time-series data and agent behaviors.

    Overview

    • Agent-Based Simulation (ABS) to generate multi-agent models.  
    • Agents can be configured with N-dimensional behaviors across time to generate trajectories of human behavior in belief space and influence networks that incorporate abstractions of Human, Social and Cultural Behavior (HSCB).
    • Valuable tool for generating simulated data for classification and clustering testing and refinement. 
    • The simulator can rapidly generate thousands of complex categorized waveforms used to train ML models.
    • As real-world data is gathered, the models are refined so that they produce accurate output. Over time, this leads to a predictive model of how individuals and groups interact with information at scale.

    Mission Application

    • The Group Polarization simulation was developed for principle research regarding trustworthy information in digital environments.
    • The data generated can be used to build supervised and unsupervised machine learning models. The characterized output data is high dimensional and time-dependent. Changes to the environment including lethal barriers and alternate behaviors can increase the variability and scope of the data generated to effectively infinite. 


     

  • Language Model Networks Icons
    Language Model Networks
    • Data science tools that augment text analytics
    Language Model Networks
    • Data science tools that augment text analytics
    Language Model Networks Icons

    Interactive Markov chain data science analysis tool for unstructured text analytics, document similarity and topic extraction

    Overview

    • Powerful data science tools that augment text analytics resulting in a ranked list of words that best represent the ideas and concepts contained in the corpus. 
    • Mines Meta-data from unstructured text (public web pages, PDF, text files) to feed downstream analytics.   
    • Based on Stanford’s Natural Language Processing (NLP) library. 
    • Statistics are collected on the lemmatized words in the document.  
    • Calculates: Bag-of-words, TF-IDF (Term Frequency–Inverse Document Frequency) and LSI (Latent Semantic Indexing) ranking of terms within and between documents.
    • Allows a data scientist to interactively reweight term/document nodes and analyze the rankings as they are dynamically recalculated.
    • The resulting output can be used to improve Information Retrieval (IR) and topic extraction.

    Mission Application

    • Successfully used since 2016 to increase commercial customers’ information retrieval accuracy of web-harvested sources by over 450%, as well as extracting topics of human behaviors to be categorized for Machine Learning classification.
    • Data science tool used when evaluating source data, by topic modeling and data tagging of textual data for analytics. This data can include log file analysis, open web content, text files or emails.
    • Unstructured flight mission event log files can be analyzed and processed for actionable information, sequencing of events or anomalies, monitoring the health of ground or flight systems, or monitoring the operational status/health of facilities.

     

  • Machine Vision Icon
    Machine Vision
    • Real-time object detection from complex images
    Machine Vision
    • Real-time object detection from complex images
    Machine Vision Icon

    Overview

    • Edge-based machine vision capability provides real-time object detection from complex images and scenes.

    Mission Application

    • Used to detect objects in streaming video or static images. By applying advanced data augmentation techniques, trained models can be developed with far fewer images than typically needed.  

     

     

  • Time Series ML Icon
    TimeSeriesML
    • Time series prediction, classification and anomaly detection library
    TimeSeriesML
    • Time series prediction, classification and anomaly detection library
    Time Series ML Icon

    Time series prediction, classification and anomaly detection library

    Overview

    • Toolkit of capabilities for classification, forecasting and anomaly detection of time series data.
    • TimeSeriesML can execute in batch model or perform live/real-time inference with appropriately defined/configured hardware or cloud capabilities.
    • Includes optimizers for hyperparameter tuning and automated model fitting.
    • Supports data visualization and plotting locally or headless via PyPlot.
    • Implemented as a Python library which includes statistical models (ETS/SARIMAX) and neural networks (RNN/LSTM) leveraging statsmodels, scikit-learn and Tensorflow.
    • Configured with Gradle/Docker to run at scale in cloud environments using GPUs when appropriate.
    • Data connectors for Hadoop Distributed File System (HDFS), Amazon Simple Storage Service (S3), and PostgreSQL DB.

    Mission Application

    • TimeSeriesML is currently deployed in batch-mode for NOAA Financial Management Data System (FMDS) running predictive analysis on the NOAA monthly financial data ingest. This instance is deployed within the NOAA FISMA boundary. 
    • Useful for any situation that includes supervised learning for time dependent data including: language analysis, telemetry data, signal processing and cyber security.
    • Automates spacecraft operations, provides anomaly detection, sensor quality assessments, and improves system resiliency for time dependent, dynamical systems. This application is deployed on NOAA GOES-R ground system/operations (since 2016).


     

  • Key Customers

    Machine Learning Key Customers Logos

    Our Team

  • Theresa cauble
    Theresa Cauble
    • AI / ML Systems Engineer
    Theresa Cauble
    • AI / ML Systems Engineer
    Theresa cauble

    Theresa Cauble is ASRC Federal’s AI/ML Systems Engineer. Cauble has an extensive history of technical experience in product development. Her ability to translate customers’ needs into executable engineering requirements ensures the resulting product features exceed expectations.

    She has extensive experience in all phases of the software product lifecycle including software development, system engineering, technical management, product-line management, technical marketing, customer advocacy, quality controls and business operations. She has the unique combination of technical expertise and communication skills to develop value-added solutions.

    Prior to joining the ASRC Federal team, Cauble spent 25+ years in the telecommunications industry at such companies as AT&T Bell Labs, Paradyne, Internet Photonics and Ciena.

  • Photo of Aaron Dant
    Aaron Dant
    • Chief Technologist
    Aaron Dant
    • Chief Technologist
    Photo of Aaron Dant

    Aaron Dant is ASRC Federal’s chief technologist. He has over 20 years’ experience building enterprise scale applications, including 10+ years developing cloud scale analytic systems for classified federal customers. He has also served as the principle software architect for numerous enterprise systems supporting petabyte+ data requirements.

    As the ASRC Federal AI/Machine Learning technical lead, Dant is responsible for developing cognitive computing solutions for commercial healthcare and federal customers using a variety of techniques including natural language processing, supervised and unsupervised learning.

    Dant's full stack programming experience, combined with his expertise in cloud architecture, big data analytics and machine learning, guides our continuous enhancements and seamless deployments.

  • John Donohue Headshot
    John Donohue
    • Senior Engineer
    John Donohue
    • Senior Engineer
    John Donohue Headshot

    John Donohue joined ASRC Federal in 2016 after a 32 year career at NASA/Goddard Space Flight Center. His career included supporting Goddard missions within the Engineering Directorate as a computer engineer and systems engineer; and then as a leader/manager within the Directorate. As a senior executive, he served as the Division Chief of the Software Engineering Division and as the Goddard Deputy Chief Information Officer. At ASRC Federal he serves as a program/project manager and senior engineer. 

    Donohue holds a Bachelor’s of Science in Electrical Engineering from the University of Maryland, and a Master’s of Engineering Management from the George Washington University.

  • Phil Feldman Headshot
    Philip Feldman
    • AI / ML Futurist
    Philip Feldman
    • AI / ML Futurist
    Phil Feldman Headshot

    Phil Feldman is ASRC Federal’s AI/ML Futurist. His most recent work has been to explore how to detect and use search patterns to determine the trustworthiness of information in scale-free and domain-independent ways. This work blends natural language processing, unsupervised learning, high-performance graphics and agent-based simulation. Phil is recognized as a thought leader in AI/ML and has written numerous peer-reviewed papers and presented at several recognized international conferences. 

    Phil has a MS in Human-Centered Computing (HCC) from UMBC, and is currently pursuing his PhD.

  • Dr. Zhenping Li Headshot
    Dr. Zhenping Li
    • Chief Technologist
    Dr. Zhenping Li
    • Chief Technologist
    Dr. Zhenping Li Headshot

    Dr. Zhenping Li is ASRC Federal's chief technologist. Dr. Li has more than 20 years of experience working on various aerospace programs that have supported numerous agencies including NASA and NOAA.  Dr. Li is an expert in applying machine learning to space missions, algorithm development in satellite instrument data processing, and developing automation software in ground systems.  Throughout his career, Dr. Li has published many articles and white papers and is well recognized in the machine learning arena. 


    Dr. Li earned his Ph.D in physics from the University of Tennessee, and holds a masters degree in Computer Science from John Hopkins University.

  • Mike Peacock Headshot
    Michael Peacock
    • Director of Innovation Engineering
    Michael Peacock
    • Director of Innovation Engineering
    Mike Peacock Headshot

    Michael Peacock is ASRC Federal's Director of Innovation Engineering. Peacock is an accomplished professional with more than 19 years of experience developing and designing high-available, real-time, mission-critical software and the utilization of Big Data Platforms, Data Analytics, and Cloud-based solutions to provide enhanced situational awareness. Peacock is responsible for supporting ASRC Federal leadership to define capabilities, technical landscape, and growth strategies, and provides development, technology, and architecture solutions for research and development activities, internal and external organizations, and customer-focused initiatives.

    Peacock has holds a Bachelor’s of Science in Chemical Engineering and Computer Science from Drexel University, and a Master’s of Computer Science from Rowan University.

  • Upcoming Events
    • ACT-IAC ELC – Philadelphia, PA – October 20-23, 2019
    • ATCA Annual – Washington, D.C. – October 20-23, 2019
    • SpaceCom – Houston, TX – November 20-21, 2019
    • AMS Annual – Boston, MA – January 6-10, 2020
    • Sea Air Space – National Harbor, MD – April 6-8, 2020
    • National Space Symposium – Colorado Springs, CO – March 30-April 2, 2020
    Want to Learn More?

    Please reach out to our team to learn more about how our AI/Machine Learning toolkit can help you achieve success and increase efficiencies.

    Contact Us