Skip to main content

Student explores facets of data science during Honeywell internship


Working alongside industry data scientists and data engineers, Di Pang was immersed in a world where he could apply his knowledge and expertise in data sciences in a new way.

Di, a Graduate Research Assistant in the WVU Lane Department of Computer Science and Electrical Engineering, is also a member of the Center for Gravitational Waves and Cosmology.

Di Pang

Over 12 weeks in the summer of 2021, Di was a Data Science Intern at Honeywell Connected Enterprise.  Honeywell International is an industrial conglomerate with areas in aerospace, industrial manufacturing, chemical and materials and building technologies.  Under two mentors from the Data Science Team, he was also able to work alongside data engineers to better understand data processing systems.

His team’s project was to detect anomalies in the Honeywell Global procurement data which are text, so the main techniques used included text mining, machine learning, and outlier detection.  The project complemented his current graduate research projects in many ways including machine learning algorithms that are used across the spectrum with some modifications.

“This enabled us to make progress the project quickly once I understood the data,” Di explains.   “One exciting moment during the internship was that I modified the Decision Tree algorithm to find an answer to a problem that was waiting to be solved for some time.”

Looking back on his experience, he can share some valuable advice.  Communication or storytelling in your research is important.  Additionally, he adds “keep learning as data can be from different domains and technology evolves fast.” 

Currently, Di’s research area is developing automatic astrophysical pulsar searching methods using machine learning, more specifically single-pulse searches in the time domain.  Research challenges that he is working on within this project include data imbalance, weak signals, and the requirement of high recall.  On these methods, Di states “we have developed a two-stage approach, named Single-Pulse Event Group Identification (SPEGID), that can automatically identify and classify single pulses.” This approach redetected many known pulsars and has made several new discoveries in pulsar astronomy. 

Di has published his research where he has presented approaches to using machine learning to identify and classify single pulses in radio pulsar search data.

His research spans two colleges and exemplifies interdisciplinary research collaboration working under his advisor Dr. Katerina Goseva-Popstojanova, also in the WVU Statler College of Engineering and MineralSciences and the Center for Gravitational Waves and Cosmology.

In the future, Di hopes to work in the industry sector as a Data Scientist or Machine Learning Engineer to solve complex and interesting problems using his data analytics skills.