Cutting through data silos to reduce unemployment

A nonprofit organization has developed two solutions to help states use data to improve their workforces.

One is affectionately known as “the pizza tracker,” said Scott Jensen, CEO of Research Improving People’s Lives (RIPL), a nonprofit that was incubated at Brown University as a policy lab. It uses cloud computing to let unemployment insurance (UI) applicants track the status of their claims, much as Domino’s web tool lets customers track their pizza orders, Jensen said.

The Rhode Island Department of Labor and Training (DLT), where Jensen was director until February, was the first to use it. During the pandemic, it partnered with RIPL to launch a pandemic unemployment assistance application in the Amazon Web Services cloud. Today it’s known as UI Online.

DLT and RIPL replicated the entire UI database, stored in an AS/400 mainframe, in the cloud using Qlik Replicate, which enables organizations to integrate data from on-premises sources to Amazon clouds.

“The source of truth is the mainframe terrestrial system, but the cloud-based system is kept synched with the on-prem system. Then we can work on the cloud to add functionality,” Jensen said. “As the waters were rising, we would take certification to the cloud, we would connect a cloud-based call center – all of these different small but strategically important interventions to get that thing onto the cloud” during the influx of UI applications last year.

That led them to think about getting people back to work when the health crisis eased, and the Data for Opportunity in Occupation Reskilling Solution was born. It uses artificial intelligence, machine learning and secure cloud computing to bring together previously siloed government administrative data, such as wage, UI and job training program records. DOORS then analyzes the data and makes suggestions about career paths and training and reskilling opportunities.

To use it, state residents sign on, answer questions about past jobs, education and skill sets and then upload a resume.

“We’ve had our team working on sophisticated topic modeling that will extract skills from your resume,” Jensen said. “A lot of topic modeling is just work matches, [so it will] do that but then adding on top of it the context of the spatial relationships in the resumes.”

As a result, users see the types of jobs that other people with the same skills hold and get suggestions about available jobs and training. They can click through to apply for a job or sign up for training. States can opt to have DOORS email users a list of results, Jensen added.

What’s more, they can learn about a new career path. That’s because DOORS also mines the state’s database containing information on the earnings of every W2-holding employee each quarter. With that data, DOORS then develops an aggregate understanding of career switches because it can see that, say, 2,000 people worked in one industry before successfully switching to another, enabling it to recommend a similar transition for people in that first industry who are now out of work.

Jensen likens it to how the Netflix streaming service suggests other shows for viewers to watch based on their watching habits and others with similar taste.

“It’s not meant to replace job coaches or people who work in one-stop systems in labor departments,” he said. “It’s meant to be a tool for people to use online or, even better, to be used in conjunction with job coaches and others.”

Next up for DOORS is an employer-facing portal. Employers will enter information about open jobs, and then the solution will search through unemployed people’s skills. When it finds a match, it alerts the employer that, for instance, Person 1234 has the qualifications for the job. If the employer wants to connect, the solution will alert the job seeker and make the introduction.

“If you can use AI to look at somebody’s qualifications and go try to find jobs that match them, you can also start with jobs and go look for people,” Jensen said.

Data sharing in Colorado

Colorado’s Department of Higher Education, Department of Labor and Employment, and Workforce Development Council are banding together to work with RIPL on a research data lake, a secure, cloud-based system that stores, anonymizes and integrates administrative data.

It will work in tandem with the state’s existing data trust. Built with BrightHive, the trust, a legal structure, maintains and manages how data is used and shared. The idea is to “create that space where we can -- with the data that are contributed to the data trust and the research data lake -- dive into various use cases and research questions, hopefully helping us better align all that we’re doing and better informing policy,” said Michael Vente, chief performance officer and senior director of research and data governance at the Department of Higher Education.

“I really am looking forward to leveraging the data that we state agencies had in better, more innovative ways to tell a more holistic story about an individual’s experience when they are interacting with various programs, and what it looks like when they … complete a higher education program of any type and then go on to the workforce,” he added.

Additionally, with the research data lake, the agencies will be able to better connect what postsecondary credentials are being awarded and what skills the workforce needs to better fill gaps.