Period of Performance: 2024 – 2025
Within the larger Satori project (combining time series analysis with crypto technologies), Dr. Soller designed the framework for supporting auto-ML (automated Machine Learning) applied to forecasting of future values of multivariate time series data. This incorporates more than 40 different computational time series analysis and statistical algorithms (integrating and extending multiple open source software frameworks), which can be divided into the following categories: baseline, classical time series, tabular/tree-based models, artificial neural networks using deep learning, Large Language Models (LLMs) customized for time series data, and intermittent forecasting models. Dr. Soller acted as a subject matter expert on the approaches to data pre-processing and time series modeling, developed portions of the software and mentored team members developing other portions of the software. The software enabled time series data pre-processing/cleaning, feature extraction, feature selection, model evaluation and selection, hyperparameter selection, model training, and various forms of testing and evaluation, and can run on CPU, GPU, and cloud environments. CogniTech is able to customize the resulting open source for projects for its future customers.
Contract Period of Performance: 2025
This customer applies AI approaches to enable the analysis of legal documents and decision support for their clients, which includes financial institutions. In one project, provided consulting and assistance in developing an architecture and software design for semantic search of databases, including analyses of best practices for deploying search engine technologies to cloud environments, and ML and other approaches for encoding text to improve search performance. In a different project, transformed legacy data into a format where it can be managed by relational databases within an Amazon cloud environment.
Subcontractor: University of Utah, Principal Investigator (PI): Jerome Soller, PhD, Period of Performance: 2022 – 2023
In this project, we developed specialized algorithms and software for the airspace management of Unmanned Aerial Systems (UAS) and other airborne vehicles (e.g., manned airplanes, helicopters). This includes algorithms and software to design virtual highways in the air, schedule aircraft using strategic deconfliction, identify anomalies, handle contingencies and coordinate between air, ground, and naval activities. The software was implemented as cloud-deployable RESTful micro-services, using Kubernetes.
Ported preliminary implementation of Calcudos™ antibiotic decision support system to a web-based application, written in the Java programming language, and a relational database. Enhanced user interface and data analytics. Supported international retrospective clinical trials. Supported analysis to enable presentation at an international infectious disease conference. Video demonstration of the software can be provided upon request.
Phase II Period of Performance: 2017 - 2020
Developed an architecture and software prototype for importing, transforming, managing, querying, analyzing, and visualizing large simulated air vehicle trajectory and performance assessment data within a Big Data environment. Developed a suite of specialized algorithms for comparing pairs of air vehicle trajectories and another suite of algorithms using innovative and specialized ML approaches.
Periods of Performance for Two Contracts: 2015 - 2020, Approximate Value of Two Contracts: $550,000
This project developed an approach, architecture, and prototype for analyzing, mining, and visualizing historical aircraft maintenance and related data. This also included the development of ontologies and a rules-based expert system to provide decision support for personnel use during a pandemic.
Period of Performance: 2015 - 2017, Value of Contracts: Approximately $360,000
Enhanced web-based software that allows healthcare researchers to search, filter, and sort electronic patient records to support research. Enabled form-based annotation of text documents to support research. Created a software interface for Natural Language Processing (NLP) tools. Hardened and integrated open source Natural Language Processing (NLP) tools for use with a VA data repository containing patient data from across the country.
Period of Performance: 2006 - 2015, Value of Contracts: Approximately $1,000,000
CogniTech developed, maintained, and supported the web-based software used by emergency responders throughout the State of Utah. This system included the web-based data collection, searches, report generation, and SOAP/XML-based web services interfaces for more than one million patient records.
Period of Performance: 2010 – 2014, Value of Contracts: Approximately $300,000
Orbital ATK (now Northrop Grumman Innovation Systems Group) developed the IMLM DAAS system to demonstrate the potential for predicting, in near real-time, an individual motor’s ability to perform its intended mission. CogniTech developed data mining/analytics capabilities that operated on missile sensor data and data associated with the manufacturing of missiles.
Period of Performance: 2006-2009, Value of Contracts: Phase I: $120,000, Phase II: $730,000
This project dealt with prognostics, risk assessment and decision support for the health of solid rocket motors. CogniTech teamed with Orbital ATK to develop physics-based models of rocket motor cumulative damage and failure. It also developed a sensor fusion framework.
Period of Performance: 2002, Value of Contract: $100,000
Developed a probabilistic inference approach for data fusion, which combined rules and Dempster Shafer theory. Applied probabilistic model to risk assessment. Combined a rules-based expert system with a MIDAS epidemiological pandemic simulation developed by SRA for the US government.
Period of Performance: 2002
Developed a web site enabling the potential licensees of rights-protected photographic images to search based on image metadata, such as subject, activity, concept, location, photographer, keywords and other criteria.
Period of Performance: 2001-2004, Cumulative Value of Two Contracts: $850,000
Developed Personal Data Assistant (PDA) software for questionnaire authoring, data collection, and data management. The questionnaires were represented using XML. The data was managed on a central server using a relational database. HIPAA assessments were developed for the technology. Usability studies were performed on the user interface and system, with standardized mental health scales/instruments, human testers, mock patients, and mock patient data.
Period of Performance: 2000
Developed software interfaces, based on the Object Management Group (OMG) CORBAmed Clinical Observation Access Service (COAS), to securely query the DoD Health Data Repository, managing patient records for members of the US armed forces.
Period of Performance: 1999
Developed prototype software for web-based application, combining the Java programming language, CORBA object interfaces, and a relational database to manage the healthcare credentialing process for healthcare practitioners.