Building Better Models

Erin Hawley tells the story of how DataRobot (Booth 1420) helped the Department of Defense integrate a system it had been struggling with for several months.

“We … took their data set in,” said Hawley, public sector vice president for DataRobot, a five-year-old machine learning automation provider headquartered in Boston. “Within 20 minutes, we had come up with a better model.”

Afterward, the DoD group was able to learn more quickly whether incoming data changed the analytic model, so data scientists could build another model to compensate.

Speed, ease of use, and price are the messages DataRobot brings to GEOINT 2017.

“We do two things,” Hawley said. “We help agencies get insights into their own data a lot faster, and we build models for you. We take your data set and find the best model for it.”

The company does so with its DataRobot platform, which provides hundreds of already-proved models and tests them on client data in competition, then generates a leaderboard on which models are listed in order of problem-solving effectiveness. Frequently the solution is a blend of different models.

“We’re looking for agencies that are looking to use data science, machine learning, or artificial intelligence to solve a mission problem,” Hawley said.

DataRobot employs 50 data scientists, she added. “Our top cases in the federal government … are people interested in cybersecurity.”

Image courtesy of DataRobot