MS Student James Sargsyan Earns Top Rankings in AgentDS Benchmark Competition

Department of Computer Science & Engineering computer science master’s student James Sargsyan achieved top placements in the AgentDS Benchmark Competition, taking first place in the healthcare domain and fourth place in the retail banking domain. The competition evaluates how well AI agents can solve realistic data science problems across six applied domains: commerce, insurance, retail banking, manufacturing, food production, and healthcare.

Because of his first-place finish in the healthcare domain, Sargsyan is eligible to submit a paper about his approaches from the competition to the International Conference on Healthcare Informatics. In his paper, he plans to examine how healthcare machine-learning systems can be aligned with human oversight in mind, emphasizing reliable evaluation practices that help practitioners understand when model predictions can be trusted.

“Hopefully, the paper gets accepted, and I'll participate in the conference, present my ideas, get some feedback, and be able to use all this experience in the real world,” Sargsyan said. 

Sargsyan chose the healthcare and retail banking domains because they offered the most compelling opportunities to look at human-AI collaboration. Healthcare, in particular, presented a unique challenge. 

“You’re dealing with data that is different across hospitals,” Sargsyan said. “There’s no uniform structure for an AI system to rely on. I wanted to see how collaboration between humans and AI could help navigate that.”

The healthcare domain consisted of three major tasks: a 30‑day readmission prediction, an emergency department cost forecast, and a discharge readiness assessment. The two main tasks in the retail banking domain were fraud detection and credit default prediction. 

Taking place over nine days, participants in the competition submitted solutions through an online interface that automatically evaluated performance and updated the live leaderboard.

“I would go to sleep at 10 p.m. in first place for the healthcare domain and then wake up and be in 10th,” Sargsyan said. “It was just constant stress — nine days of competing  for the leaderboard score.”

Despite the pressure, the experience was rewarding.

“It was really nice to see everything I’ve learned about computer science have real results and show that I’m able to tackle real-world problems. It was also rewarding to get to know more people, their ideas, and how they approached the same challenges.”

Sargsyan’s interest in computer science began with video games and how they worked behind the scenes. In high school, he debated between pre‑med and computer science.

“There was a moment where I realized if something happens to a patient when I'm a doctor, I might take it too personally. I started thinking about how I could contribute to the healthcare field in a different way, and that led me to computer science. Computer science felt like a way to build tools that can support people in doing their jobs.”

After completing his master’s degree, Sargsyan plans to pursue a PhD in computer science, focusing on AI for healthcare. Afterward, he hopes to work in industry while teaching parttime. Sargsyan encourages students to embrace experimentation and failure.

“Don't be scared to fail. Don't be scared to try new things. Just like many computer science models improve through trial and error, you have to go through failures too. University is one of the best places to learn from your mistakes and keep improving.”