Eric kolve, Director of data science at dexcare
PLEASE DESCRIBE WHERE YOU WORK AND YOUR POSITION THERE
I work at DexCare, and I’m the Director of Data Science.
FOR WHAT PROJECTS/SERVICES DID YOUR COMPANY HIRE XYONIX?
I wasn’t involved in hiring Xyonix, but rather came in after they’d made quite a bit of progress. The projects they worked on involved developing a parser that would take a patient’s reason for visit and emit some number of labels related to healthcare. They also developed a provider match model, which would provide a prediction for how long a patient’s visit would take for a virtual visit as well as a score indicating how well the provider would be suited for that patient’s reason for visit.
HOW DID YOU SELECT THIS VENDOR AND WHAT WERE THE DECIDING FACTORS?
Unfortunately I wasn’t involved, but our CEO had used Xyonix in the past and was delighted with their work, so he selected them again.
DESCRIBE THE PROJECT IN DETAIL AND WALK THROUGH THEIR SERVICE PACKAGE.
Xyonix analyzed the “reason for visits” that we received from our patients and matched them to different health systems that we provide service to. The “reason for visit” ranges from a short-one sentence description to a paragraph describing the patient’s condition. Xyonix examined tens of thousands of these reasons for visit, and developed a custom ontology to describe the types of things people come in with. We ended up with 150 distinct labels, anything from “COVID-related” to “flu-related” to “needs prescription”, etc. Once they arrived at this ontology, Xyonix trained in-house labellers to review the reason for visits after performing an integration with their own database, and applied these labels. Each reason for visit had n number of labels attached to them, which became ground truth for the model or parser. Once these ground truth labels were applied, a parser was then trained up and shipped to us in a Docker container so that we could then deploy it in our own Kubernetes cluster.
HOW MANY RESOURCES FROM THE VENDOR'S TEAM WORKED WITH YOU, AND WHAT WERE THEIR POSITIONS?
I don’t know how much Deep was involved in the development of it, but he would attend our status meetings. Carsten and Andy were heavily involved. Then I believe there were many resources that we didn’t have direct contact with who performed the annotations.
HOW EFFECTIVE WAS THE WORKFLOW BETWEEN YOUR TEAM AND XYONIX?
I was very happy with the workflow between my team and Xyonix, there was always quick communication back-and-forth over email if I had questions between status meetings. When we did have our status meetings, it was always really productive. We got a quick recap on what the status of everything was, and updates on how things were going to proceed in the future. I always found the status meetings very helpful, they offered a good chance to also brainstorm new directions we could go with our data.
WHAT DID YOU FIND MOST IMPRESSIVE OR UNIQUE ABOUT XYONIX?
Xyonix’s ability to take a problem that has very little structure around it, build an ontology, and provide an end-to-end solution was pretty impressive. You have to bring quite a few different skill sets to the table, and to have one group of people be able to do that in a really clean and consistent way was very nice and well-executed.
ARE THERE ANY AREAS FOR IMPROVEMENT OR SOMETHING THEY COULD HAVE DONE DIFFERENTLY?
I don’t think so.
HOW WAS THE QUALITY OF XYONIX’S WORK?
5/5
HOW WAS SCHEDULING WITH XYONIX?
5/5
HOW WAS THE COST OF XYONIX’S WORK?
5/5
HOW LIKELY ARE YOU TO REFER XYONIX TO A FRIEND?
5/5
GIVE XYONIX AN OVERALL RATING.
5/5