Page 25 - FCW, November, December 2018

PROJECT: Administrative Data Research Facility
Census Bureau
Data mashups at
government scale
The Census Bureau — the government’s original data agency — collaborated with the University of Chicago, University of Maryland and New York University to provide a secure cloud-based platform that allows government employees and academic researchers to take advantage of advanced data science tools.
The goal of the Administrative Data Research Facility is to give authorized
users access to a secure supercomputer and thousands of datasets. It has received a Federal Risk and Authorization Management Program moderate certification.
John Abowd, an associate director
and chief scientist at the bureau, said it
is the first such platform in the federal government. “The ADRF itself has done
a huge service by providing a usable computing environment that meets [different users’] specific conditions, and [it] has the data science tools” for a range of projects, he added. In that way, “the Census Bureau made the investment on behalf of other agencies.”
The platform is being tested by users from multiple government organizations and so far hosts almost 50 confidential datasets from 12 agencies, supporting a variety of potential uses, Abowd said.
“Another government agency might want to do a program evaluation [and] combine its data with a cooperating agency, whether or not that analysis is something the Census Bureau wants to do,” he said. “The ADRF was designed for that.”
The platform has also helped the Census Bureau refine its ability to incorporate functionality and security into future projects. “The facility helped us streamline the way we get a cloud service properly secure to meet the standards of a Census Bureau [authority to operate],” Abowd said.
BEST IN CLASS - FEDERAL CIVILIAN
PROJECT: BuySmarter Initiative Department of Health and Human Services
AI and automation to transform acquisition
At the Department of Health and Human Services, acquisition is a complex beast. The agency spends about $24 billion annually on a wide range of items, from cutting-edge technology to nitrile
gloves. And 39 separate acquisition offices
bring their own labor-intensive processes to the task.
That approach is inefficient and
expensive. HHS officials have discovered
price variations of more than 300 percent
for the same products purchased under
identical terms and conditions. To improve
the process, the agency developed the
BuySmarter initiative, which capitalizes
on machine learning, robotic process automation and blockchain ledger technology.
The initiative started as a proof of concept to ingest, structure, analyze and report on data across HHS. Using IBM’s Watson, BuySmarter learned from
18 months’ worth of contracting data and now enables acquisition teams to access real-time information on departmentwide pricing and the terms and conditions for 10 categories of purchases.
The potential payoffs are massive. Lori Ruderman, BuySmarter project lead, said her team’s conservative estimate is that extending the best prices across HHS would result in $720 million in annual cost savings.
Additionally, all contracting data is now stored in a single database that is structured according to federal category management guidelines. There is also the potential for the project to be extended to other agencies.
Although HHS officials are not planning to offer a shared service, the reliance on a blockchain data layer and connecting microservices means
it would be relatively easy for other agencies to join the ecosystem, said Jose Arrieta, associate deputy assistant secretary for acquisition at HHS. “Especially if you’re a smaller agency, wouldn’t you want instant access to $24 billion worth of purchasing data?” he asked.
November/December 2018 FCW.COM 25