Marquette Students Embrace Research Opportunities at SLU, Earn Chance to Present at National Conference

Marquette High seniors Adithya Chengalvala and Satvik Seetharaman both benefited from valuable research opportunities in machine learning at Saint Louis University (SLU) this past summer through the Rockwood Partners in Education (PIE) Summer Internship program.

Now, both will also get the opportunity to help present their teams’ research at a national geospatial conference in December.

Chengalvala and Seetharaman’s research projects have both been accepted as part of the American Geophysical Union (AGU) Bright STaRS (Students Training as Research Scientists) program, giving them the chance to present to the approximately 30,000 scientists attending the annual AGU conference, which runs this year from Dec. 15-19 in New Orleans, Louisiana.

“I feel incredibly grateful that I was given the opportunity to conduct guided research with extremely helpful peers, as I learned a lot more technical coding skills and also greatly improved my ability to communicate research with others,” Seetharaman said. “This conference is an amazing opportunity for me to learn how to present research effectively and get advice from professionals in the field. It will also expose me to other research being done, allowing me to learn how I can continue to make contributions to society through applied computer science research.”

The two Rockwood students were matched with Dr. Orhun Aydin, SLU assistant professor of earth and atmospheric sciences and computer science, and his AI-CHESS (Artificial Intelligence in Coupled Human-Environment Systems for Sustainability and Resilience) Lab through the PIE summer internship program, which provides students with career exploration opportunities in a field of interest.

Both students used aspects of computer science and machine learning to help with ongoing projects in the AI-CHESS Lab. Chengalvala studied tree placement to help mitigate urban heat islands, and Seetharaman studied identifying different types of waste deposited in recycle bins.

“Tech can be used to solve so many problems, and tomorrow it could be used for something different that we can’t even imagine right now,” Chengalvala said. “We can find ways to give life to these machines by integrating computer science, and that’s what where my passion for computer science comes in. I can create something that can help someone’s life become a lot better.”

The urban heat island effect explains the process by which – through city surfaces such as buildings and pavement that retain heat – urban centers are often considerably hotter than surrounding areas.

Aydin’s lab is researching ways in which the placement of green areas in cities can be optimized to help mitigate heat islands. Chengalvala’s part of the project deals with teaching a computer model to be able to identify where trees are placed in urban settings through a mixture of satellite and aerial imagery.

Both types of imagery have their drawbacks: satellite does not have the appropriate resolution for accurate machine identification of trees, and aerial lacks important data – such as object height and crown width – that satellite carries.

So Chengalvala’s model seeks to combine the two inputs into one output.

“This is the premise of the project in its simplicity … obviously on top of thousands of lines of code,” he said, with a laugh. “How do we develop urban districts in such a way that we maximize the cooling effect of trees and grass while also having the basic amenities of a city? I had to find ways to incorporate aerial imagery, which is easier to look at, while still getting the data out of satellite imagery.”

In Seetharaman’s project, he wrote code for three convolutional neural networks – which use three-dimensional data to for image classification and object recognition tasks – to help determine the types of waste in a recycling bin under a variety of different lighting conditions.

That way, researchers can see which models work better in which conditions and try to optimize a model to fit best in all conditions.

“I've been greatly interested in computer science over the past few years, working on multiple independent research projects in bioinformatics, computational game theory and machine learning in the criminal justice system,” said Seetharaman, who also competed in the Regeneron International Science and Engineering Fair in Columbus, Ohio, in May. “This has inspired me to continue doing computational research that can truly make a difference in our world. Working at the AI-CHESS Lab gave me a chance to apply my proficiency in computer science to work in a professional and collaborative setting, which turned out to be incredibly useful.”