4 minute read

A SOURCE OF OPPORTUNITY

By Katie Ray

In an effort to provide opportunities for employers to fill gaps in their workforce needs, Eastern Kentucky Concentrated Employment Program launched a new initiative, the One Up internship. One Up works to match the skills needs of area employers with the career interests of current college students and provide them with real-world learning opportunities and the potential for employment if a solid match is made in an internship experience.

As part of One Up’s pilot program, Pike County native, UPIKE junior and mathematics major Ashley Harrison was one of two participants who recently completed internships at Bit Source, a tech company that specializes in software development located in Pikeville.

“There is not a clear path from a mathematics major to a specific career, but there are plenty of opportunities for math majors, outside of education,” said Harrison. “I wanted to explore careers in various industries that could use my knowledge of mathematics, and I know data science involves a lot of math, so I was excited to give it a try.”

Harrison says all her teachers have been influential to her to some degree, but her academic advisor UPIKE Assistant Professor of Mathematics Bang Huang, Ph.D., has helped guide her tremendously.

“Ashley is an outstanding student. She is dedicated and hardworking. She had a solid mathematical background before she came to UPIKE, but she was still working hard in all her classes,” said Huang. “She loves to help other students get better, and she is good at explaining things to them. Ashley is one of the best students, and the type that all the professors want to have in their class.”

UPIKE Professor of Mathematics Michael Holcomb, Ph.D., says Harrison has a vast understanding of both computational and conceptual mathematics, even in the more complicated topics.

“She has a keen eye for detail and submits high-quality and organized assignments,” said Holcomb. “I have enjoyed watching her grow academically and personally in the math major.”

In addition to Bit Source, Co-Founder Rusty Justice also co-founded Jigsaw Enterprises, LLC, an applied engineering company. He says they saw a need to make their concrete inspection more efficient.

“We have a team that performs concrete inspections for Pikeville, and right now, those inspections take a long time to do and are labor intensive,” said Justice. “We saw an opportunity to automate this and bring down the cost to grow this as a business and a service that we can provide not only locally, but across the country.”

The internship began in June and ended in mid-August. Alongside Evan Hamilton, who is a senior computer engineering student at the University of Kentucky, Harrison worked on machine learning to improve the efficiency of concrete inspections.

Machine learning is the study of computer algorithms that can improve automatically through experience and by the use of data. It is seen as a part of artificial intelligence.

“I worked to optimize the performance of a Convolutional Neural Network for concrete damage inspection, which is basically image classification,” explained Harrison. “We collected images from the Pikeville and Paintsville areas and sorted them into four classes of damage that range in severity.”

Images were collected with an iPhone and the team used the data they recorded to teach the machine to analyze pixel information into groups and recognize patterns to automatically sort the photos into the categories necessary.

Harrison analyzed the model to discuss and present findings within data iteration of the entire training dataset the machine learning algorithm has completed. The model is trained and re-trained on this set of data and it continues to learn features from the data in each set.

“I worked to expand our training and testing sets through data augmentation. I tuned the hyperparameters based on cases of the model underfitting or overfitting our data,” said Harrison. “I provided insight on machine learning processes like convolution, backpropagation and a variety of loss or activation functions.”

Bit Source Chief Operations Officer and Partner Payton May says there are a lot of different machine learning models out there, but each of them deploys a variety of mathematical principles to analyze data and look for patterns.

“Ultimately, it learns from the patterns it’s spotting. Utilizing machine learning and photo data to create models that can accurately predict and classify concrete issues into the four classes will cut down on a lot of time,” said May. “Ashley had a very successful presentation and several models with significant progress toward a final product.”

The model achieved a stable accuracy range of 75-80 percent.

“The data scientist we worked with said we should aim for 85 percent accuracy to prevent overfitting,” said Harrison. “We also had minimal loss.”

To Justice’s surprise, Harrison was unaware there was an opportunity for this type of work so close to home.

“Our goal is to reimagine our local area and create opportunities so our people can stay here. We need our best and brightest to stay local,” said Justice. “Ashley was an ideal intern. She was disciplined and curious, very curious, which is a trait we value. She was friendly, kind and very intelligent. She was your typical hillbilly, which is something we truly admire.”

The next steps are to collect more data for the model to train on and deploy the model into production with GPS technology. The project will be continued by a Bit Source and Jigsaw Enterprises team, and hopefully, one day commercialized.