Capability: Artificial Intelligence and Machine Learning
Gunnison’s AI & machine learning engineering approaches enable superior data and record linkage solutions in the Federal space. These techniques, based on complex statistical evaluations, enable disparate matching of people and places across diverse datasets throughout the United States. Our core capabilities include:
- Data and Record Linkage
- Leverage High Performance Computing Environments (HPCE)
- Predictive Modelling
- Natural Language Processing (NLP)
- Distributed Processing
- Computational Intelligence
- Entity Resolution
- Distributed Bayesian Linkage
Recent Gunnison AI Projects
The United States Census Bureau Center for Optimization and Data Science (CODS)
- Gunnison developed an entity resolution probabilistic match solution for the Citizen Voting Age Population (CVAP) program to clean survey data because of a lack in deterministic and unique identifiers.
- Our methodology combines unsupervised machine learning algorithms to resolve the entities by searching for clusters of records that have sufficiently similar attributes.
American Community Survey Office (ACS)
- Gunnison leverages its unique ML capabilities through the ACS Contact History Instrument (CHI) text analytics project.
- Gunnison developed a combination of supervised ML models using CHI text data and case notes. The goal is to better understand drivers of movement, based on CHI and case notes using Response Propensity Models, Burden Score Analysis, and Markov State Space Models.
U.S. Patent and Trademark Office Software Quality Assurance Division (SQAD) Testing Services
- In order to move faster through the regression test cycles while supporting continuous deployments under DevOps at USPTO, Gunnison introduced AI to reduce the test automation code maintenance and improve efficiency.
- Gunnison proposed the use of ‘Self-Healing’ AI features which prevent script failure. Our AI algorithms can scan through system objects and provide intelligent alternative properties to the failed object.
- By introducing AI to our automated testing approach, we have the capability to validate trademark images, scanned documents, and other PDF content which may be submitted to the USPTO during the patent approval process.