Driving Decisions with Data Science

ADIApproved

30 Oct 2017

At Maxim, a new Machine Learning Club is bringing together experts from different business units to share their knowledge, solve problems, and find more efficient ways to get things done. The company also has a data science team that works to enable data-driven decision-making across all business processes.

"Employees have many skills, but in their job, they may not need to use all of them. The Machine Learning Club brings together people from different departments with data science expertise to find ways to solve problems," explained Serdar Kizilgul, a principal member of the technical staff who helped form the club. "We challenge each other."

Indeed, numbers don’t lie, and data analysis helps to weave them into a story. Many industries can benefit from the patterns and insights uncovered by data science. Traditional businesses are diving in (note Ford’s $1 billion acquisition of Argo AI.) Even semiconductors are no exception when it comes to the benefits of machine learning. The growing demand for data science in workplaces is fueling universities to add relevant programs. UC Berkeley, for example, is creating a new data science undergraduate major in response to the soaring popularity of its first Foundations in Data Science course. As Forbes notes in a machine learning article earlier this year, "Through the accessibility of cloud computing, the ubiquitous availability of parallel processing power, near free data storage and an exponential increase in data input from consumers, we have reached a perfect storm that is enabling artificial intelligence to truly have its day in the sun."

Chip design is one of many industries that can benefit from machine learning.

Seeking Greater Accuracy in Cuffless Blood Pressure Measurement

Maxim’s Machine Learning Club recently hosted a competition exploring ways to achieve accurate cuffless blood pressure measurement. Without the traditional cuff, blood pressure measurement can be performed on a fingertip. However, fingertip measurements involve addressing the effects of noise. For example, simply pushing down on a finger creates changes in blood pressure, making an accurate reading difficult. This is a challenge that many companies are working on, and one that will need machine learning and data science to solve.

The Machine Learning Club started with a blood pressure trending data set that had been validated by an external lab via an independent clinical study. This data set was used to tune Maxim’s blood pressure measurement algorithm. Nearly 30 competitors experimented with different models to increase measurement accuracy, ensuring that measurement errors would fall within tolerances allowed by internationally accepted blood pressure measurement standards. Together, the participants’ efforts further refined Maxim’s blood pressure algorithm. There are plans in the works to license this algorithm to customers in the future.

Building Problem-Solving Algorithms

Yakov Shkolnikov, one of Maxim's data scientists, placed first in the club's competition. Shkolnikov, who has done work previously on medical devices, brought to the effort insights on clinical design and algorithms. For his day job, he noted that he and his team typically get involved in data analytics projects that either help improve the bottom line or generate revenue for the company. Formed in 2014, the company's data analytics team stemmed from a data summit where representatives from different companies discussed how they utilize data science. Today, Maxim's data analytics team supports new product development, helps optimize or define new processes, provides forecasting and demand planning, and much more. The team's vision is to continue to enhance the company's data repositories, help establish critical connections between data sets, and enable improved decision-making throughout the Maxim enterprise, said Bill Hanley, executive director of data science.

Meanwhile, the Machine Learning Club will continue to be a forum for employees to share their challenges with data experts who can together come up with solutions.