Production Perig - stock.adobe.c
When COVID-19 began to spread in early 2020, Vyaire Medical needed to react.
And react quickly.
Vyaire wasn't affected by the pandemic the same way most enterprises were when stay-at-home mandates led to an economic downturn. The company wasn't suddenly struggling to survive, and demand for its products didn't evaporate seemingly overnight.
Vyaire makes ventilators and other respiratory equipment, and abruptly, as people started getting sick, ventilators were in desperate need.
If Vyaire didn't respond by increasing production -- substantially -- it would cost lives.
By sheer coincidence the company, founded in 2016 and based in Mettawa, Ill., had begun to develop a data-driven culture in early 2019, and the tools it had only recently started using from AWS to bring together siloed data and migrate its data to the cloud to develop a single source of analytics information were suddenly critical guides as Vyaire reacted to exponential demand growth.
On a normal day before March 2020, Vyaire completed building six ventilators per day. Once the pandemic hit, it needed to make 750 per day.
"We had to quickly ramp up," said Gopal Ramamurthi, senior director of the analytics and enterprise data platform at Vyaire, during a breakout session at AWS re:Invent, the tech giant's virtual user conference on Dec. 2. "To put it in perspective, it wasn't about whether we wanted to take on this business or not; it was about saving lives, so we had to get it done."
About a year before the start of the pandemic, Vyaire launched what it termed Project Insight in concert with AWS in early 2019. Project Insight, simply, was Vyaire's transformation from nearly 12 disparate data sources -- the company has grown through acquisitions -- to one analytics platform, and from an ad hoc decision-making philosophy to a data-driven culture.
The company adopted Amazon Redshift as its database, Amazon QuickSight as its business intelligence platform and Amazon Elasticsearch as its search and analytics engine.
And once COVID-19 began to spread and demand for ventilators increased exponentially, the efficiency and level of insight Vyaire was able to develop with analytics became critical to ramping up production and meeting demand.
"Vyaire was able to improve manufacturing throughput by leveraging data captured in Project Insight as part of a machine learning model, [and] used analytics and machine learning tools on AWS to change decision-making and rapidly scale its production capabilities in the face of a monumental increase in demand," said Joshua Kahn, principal solutions architect at AWS, during the virtual session.
The first thing Vyaire had to do to meet demand was increase its manufacturing space and staff, and the company accomplished that by forming a joint venture with Spirit AeroSystems.
Once it had the space and personnel in place, the actual task of producing 750 new ventilators per day begun.
"This is where analytics services came into play," Ramamurthi said.
Vyaire began by examining its supply chain to determine how many raw materials it had on hand, how much it would have to increase order volume to meet its new demand, and which vendors it could use to reliably get the parts it needed to develop 750 ventilators per day. It also had to discover where the end users of the ventilators -- the hospitals -- were located and how it would get the ventilators to those medical facilities.
Gopal RamamurthiSenior director of the analytics and enterprise data platform, Vyaire Medical
And analytics was also the means by which Vyaire developed its production plan, one that Ramamurthi said is constantly changing as the number of COVID-19 cases rise and fall and the degree to which the vendors on which Vyaire relies are able or unable to meet its needs.
According to Ramamurthi, Vyaire's supply chain data updates every 15 minutes so the vendor can react to any changes in real time.
Another key step was developing machine learning models to improve the failure rate of Vyaire's ventilators during the production process.
Ramamurthi said that Vyaire tests its ventilators three times during production, and prior to the pandemic about 73% of the ventilators failed at some point and had to be reworked to ensure that they'd work when needed to save a life. That failure rate was acceptable when only six ventilators per day were being made, but when 750 per day were in production that rate was simply too high given the labor needed to discover the problem and the time needed to fix it.
Using Amazon SageMaker, Vyaire built and trained machine learning models that helped predict failure and enable prevention of failures. The result was a significant reduction in the failure rate.
One more step Vyaire took to respond to the increase in demand due to COVID-19 was to hold a daily production review looking at key performance indicators.
"There is still tons to be done internally," Ramamurthi said. "On the business process side, there are several areas that need attention."
To that end, Ramamurthi said Vyaire plans to start using AWS Glue to help extract, transform and load its data, and take further advantage of SageMaker to better predict demand forecasting and product availability.
"We are now truly data-driven, from what was chaos before," Ramamurthi said, quoting Vyaire CIO Ed Rybicki.