Student Spotlight: Co-op Tackles Sports Tech Challenge with Computer Vision

When player statistics and game analytics appear on his TV screen during the FIFA World Cup, Raj Kavathekar sees more than broadcast graphics and data. A Master of Science in Computer Science student at Northeastern University in Miami, Kavathekar spent his co-op working on a sports problem that lacks an established solution: generating live performance metrics using AI.

The opportunity also gave Kavathekar firsthand experience in an AI startup environment — an example of the kind of early-stage work that is increasingly part of Miami’s tech ecosystem.

“My co-op did more than expose me to sports analytics,” Kavathekar says. “It showed me how many different directions AI can take you.”

Journey to a Master’s in Computer Science at Northeastern University

Before arriving at Northeastern University in Miami, Kavathekar thought he had a clear idea of where his career was headed. He earned his bachelor’s degree in game and interactive media design and worked as an AI designer. But over time, Kavathekar realized that his skill set was narrowly focused.

“A lot of master’s students go to grad school to get a specialized degree, but for me, it was the opposite,” he says. “I was looking to extend my knowledge and gain a broader foundation. I knew I wanted to stay in AI, but I wanted to open my career path beyond gaming.”

Having completed his undergraduate education in Orlando, Kavathekar says Florida already felt like home. So, when he began to explore graduate programs, Northeastern University’s reputation and emphasis on experiential learning made it a natural fit to continue his studies in the state. A scholarship solidified the decision.

Once in the program, offered within the Khoury College of Computer Sciences, Kavathekar deliberately focused on taking AI-related courses. One course, Pattern Recognition and Computer Vision, became a defining moment.

“That course is math and physics-heavy, and it made me develop a passion for computer vision,” he says. “A lot of what I learned in class and my raw lecture notes ended up helping me during my co-op. One example was Kalman filtering, which calculates the velocity of an object. I used that to help the system I built maintain player tracking when athletes overlapped or moved fast.”

Tackling a Problem No One Has Solved Yet

Motivated to apply what he was learning in his classes, Kavathekar began attending local events and hackathons across Miami’s growing startup ecosystem. At an event at the LAB Miami, he met the founder of EasyChamp, a startup specializing in AI-powered sports competition and league management. The company helps organizers automate bracket generation, performance tracking, and statistical analysis.

That meeting put Kavathekar’s co-op in motion.

As a part-time AI engineer at EasyChamp, Kavathekar was responsible for building and testing computer vision models capable of identifying players, tracking movement, and generating soccer performance data during live broadcasts.

“If you watch sports like football or hockey, there are entire teams of analysts tracking player statistics in real time,” he says. “The goal was to bring some of that capability to smaller teams that don’t have the resources. The challenge is that when players have similar builds or skin tones, are moving fast, and overlapping, AI loses its ability to track players accurately.”

The startup environment offered an opportunity to be comfortable with ambiguity and gain cross-functional experience. Kavathekar balanced graduate coursework alongside startup demands. Meanwhile, the project itself evolved continuously as new challenges emerged.

Building the entire pipeline from the ground up involved testing new approaches, refining model inputs, and constantly rethinking how the system handled the challenge of sorting players. As he dug deeper into the problem, Kavathekar realized that current computer vision technology was not yet reliable enough to generate accurate metrics from live soccer footage. Rather than abandoning the project, he helped develop a solution that processed recorded video.

“I tried so many approaches, I was researching every day, going through class notes, and keeping up with new technology, and there were moments where I felt totally stuck,” he says. “But it taught me to be accountable for my work and be proactive. I couldn’t just say, ‘This isn’t working’ because I was responsible for the whole thing.”

A breakthrough came when Kavathekar integrated a newly released computer vision model, which significantly improved performance. Processing time for a 30-second clip dropped from roughly an hour to about 15 minutes, and the system reached approximately 81 percent accuracy.

From AI Co-op to Professional Growth

By the end of his co-op, Kavathekar says he realized his experience was less about building a single system and more about understanding what it takes to build AI in a startup.

Combined with the foundation he developed through the Master of Science in Computer Science program, the unique experience reinforced the reason Kavathekar pursued a master’s degree in the first place: to expand his options beyond a single career path. While he remains interested in computer vision and applied machine learning, he is now considering where his skills can have the greatest impact.

“Coming from games, I had a very one-track mind,” Kavathekar says. “During the program, I saw my brain expand in real time, and the co-op was such a memorable experience because it helped me think more intentionally about exploring all areas of AI to find the right fit for me.”

By: Izabela Shubair