*This article is a part of the series “Artificial Intelligence and Its Applications: Perspectives From Across Kent State,” highlighting the applications of AI in different fields and including insights from students and faculty. Stay tuned for future articles covering topics such as healthcare, translation and design.

, who earned his undergraduate and master’s degrees in computer science at , is now a third-year doctoral student and adjunct instructor researching the practical applications of neuromorphic AI models and transformer-based large language models (LLMs).
Law Enforcement
On the neuromorphic side - which is inspired by biology and the human brain - Hood’s research is based on the work of and its technology. SoundThinking, a public safety technology company, deploys AI-powered gunshot detection devices in more than 170 cities, communities, counties and university campuses.
“Think of a city like New York deploying these on a bunch of rooftops and then trying to figure out, 'Oh, we have a report of a gunshot in the Bronx. What does that mean? Where is that?'” Hood said. “They can look at this device and go, it should be somewhere around this apartment building or this convenience store.”
Hood’s research is focused on improving the efficiency of this technology by lowering the cost of running the devices and the processing requirements of the system itself.
“The model that they’re using right now for detecting these is very heavy,” Hood said. “It takes a lot of processing power and electricity, so I’m trying to find ways to make that more efficient.”
Manufacturing
On the transformer-based LLM side, Hood works with an AI research laboratory . For example, in vehicle welding, he said this is a less wasteful and disruptive way to test the welds compared to taking cars apart and checking for errors manually.
“I was interested in doing non-intrusive fault detection,” Hood said. “So that means you’re not destroying anything and you’re not altering the process of these welders in any way.”
The process involves seeing how much power a welder is drawing from an outlet in the wall and feeding that through a transformer-based architecture that is similar to an LLM, but on a smaller scale. Hood has also used echo state networks (ESNs) in this process.
ESNs are a special kind of neural network that don't require heavy training because they keep a memory-like "echo" of previous data, making them great at detecting patterns and making predictions.
"An ESN can recognize a cat after seeing just one, while a transformer might need to see 10,000," Hood said. "But once trained, the ESN only knows 'cat' or 'not-cat,' while a transformer can learn to recognize cats, dogs, buses, towns - anything, really."
In both research endeavors, Hood is applying theoretically well-documented concepts and seeing where they are applicable in a practical sense. He said ESNs are very well-documented theoretical concepts but aren’t popular in AI communities. They aren't widely used because they have limited adaptability and are less flexible than transformer-based LLMs, but are useful in low-data situations when all the data is fairly similar. Transformer-based LLMs are popular, but have historically been used for applications such as natural language and image processing.
Hood has had his research accepted in three publications focused on AI. He said getting accepted into journals and conferences and seeing his work be useful to other people is fulfilling.
“Even if it’s not directly related [to their field], as long as people are finding my work and presentations useful, that’s very rewarding,” Hood said. “The acceptance into conferences and journals gets you out there and validates your work as novel.”
Hood recommended that students who want to start studying AI should just begin the process without worrying about being perfect. He also suggested getting into linear algebra to gain a thorough understanding of how the AI models work.
“For AI, if you’re working on a problem or dataset, just find out what model you need and how to train it to get good results,” Hood said. “Then worry about how you’re going to set up the things around it and get it to work.”
Hood said he is thankful for the support he receives from the Department of Computer Science and the AI research laboratory to continue his education and research. He hopes to graduate next year and work for the AI research laboratory full time. His long-term goal is to work for a university and earn tenure.
This series of articles only scratches the surface of the applications of AI - we’d love to hear from you! Fill out the following survey to share how you use AI in your work:
###