USC Viterbi PhD Student Builds a Safer, More Reliable Future for AI and Robotics

For Yusuf Umut Ciftci, the path to USC was driven by a curiosity about how engineering and technology can solve complex, real-world challenges.

Originally from Turkey, Ciftci completed his undergraduate degree in electrical and electronics engineering at Bilkent University in Ankara, where he developed a passion for signal processing and intelligent systems. After graduating, he continued his academic journey at USC, joining the PhD program in Electrical and Computer Engineering at the USC Viterbi School of Engineering in 2021.

“Overall, USC was very helpful in my transition,” he said. “The department, the USC Viterbi School of Engineering and the Office of International Services helped me at every step of this journey.”

Currently, his research focuses on one of the most pressing issues facing artificial intelligence today: developing systems that are safer and more reliable as they are implemented and integrated throughout daily life.

“With these systems being deployed, there is a high bar on safety,” Ciftci told the USC Office of International Services (OIS). “You don’t want these agents to do any harm, but you also want them to perform the task they are supposed to do repeatedly, with little to no failures.”

From autonomous vehicles to robotic systems, Ciftci’s work examines how AI systems learn from data and how researchers can improve the information used to train them. According to Ciftci, the quality and structure of training data play a critical role in how AI systems behave.

“It is important to understand how we should collect and analyze this data, what we want in our training data and what we don’t want,” he said.

Beyond his research at USC, Ciftci recently completed an internship at the Honda Research Institute, where he worked with the mobility team on autonomous vehicle research. There, he contributed to the development of predictive planners — AI systems that help autonomous vehicles navigate complex environments.

By including examples of safety-critical situations in training data (such as avoiding collisions or recovering from dangerous scenarios), researchers can help AI systems develop safer and more effective behaviors.

“In real-world driving scenarios, it requires reasoning about many things, like pedestrians, other cars, their intentions and what the vehicle’s goal is,” he said.

The USC Viterbi PhD program also allowed Ciftci to collaborate with students and researchers in other robotics-related fields. Through a robotics seminar connecting faculty and students from electrical and computer engineering, computer science and mechanical engineering, Ciftci found opportunities to exchange ideas and develop new collaborations.

Those connections contributed to research projects exploring how AI systems can become more reliable. In one project, Ciftci worked with a group of fellow Trojans and researchers at Virginia Tech to improve how learning-based AI agents respond to complex tasks.

The team tested their approach using a robotic arm that played air hockey with itself — a challenging task requiring precision, adaptability and reliability. The project demonstrated how combining traditional engineering methods with modern AI approaches can help create more stable, dependable systems.

As he continues his doctoral studies, Ciftci hopes his research will contribute to AI systems that positively impact people everywhere.

“I would love to see my methods improving systems that are touching people’s lives all over the world,” he said.

Looking ahead, he hopes to continue pursuing research in robotics and artificial intelligence.

“It is a really exciting time for robotics,” Ciftci said. “There are a lot of unknowns, a lot of room for improvement and a lot of room for impact.”