Using Human Judgment to Advance AI

If you've ever generated an image using ChatGPT or AI, the results may not have been exactly what you were expecting. Almost getting it right might be good enough for a photo, but if you're using AI to build an electrical circuit, accuracy is the difference between success and failure. 

Enter Impromptu, an end-to-end, prompt-to-hardware circuit prototyping platform that automates the design, verification, and assembly of electronic circuits. The project, published in IEEExplore and presented at the 2026 IEEE International Symposium on Robotic Sensors Environments in Norfolk, VA, began as an electrical and computer engineering Senior Design Capstone effort.

"We wanted to do something difficult and with interesting impacts," says team member Vincent Alesi. "AI is so big, but it isn't nearly as good as it could be. No matter how advanced the system gets, it can't replace human judgment.

"Our software and hardware link, using circuit analysis and verification, providing a workflow that's as easy as typing a prompt to a chatbot, with the real-world circuit-building impact of an electrical engineer at your fingertips."

Alesi, along with teammates Absa Fall, Matthew Grimalovsky, and Adrian Jackson under the supervision of Daniel Burbano Lombana, electrical and computer engineering assistant professor and Sasan Haghani, teaching professor and capstone coordinator, developed a system in which a user provides a natural-language description of the desired circuit. A system of large language model (LLM) agents then generates a candidate circuit design, which is checked through SPICE-based simulation, before the system sends assembly instructions to a robotic Cartesian gantry. The gantry physically places components and connections on a breadboard. 

"The goal was not simply to have an LLM suggest a circuit," explains Alesi. "We wanted to create a workflow in which the proposed design is computationally verified and then translated into a physical prototype."

This prototype system, the team believes, is the first of its kind to integrate software with hardware that serves as its own troubleshooter, removing the guesswork from a project. 

In addition to reducing errors, Impromptu is quick—input to output is about two minutes—and brings the expertise of an electrical engineer into the room, wherever that might be worldwide.

"Our capstone project felt like the perfect way to bring my Rutgers journey full circle," says Fall, who transferred to SoE in her sophomore year. "A new campus, new professors—it wasn't always easy, but the experience pushed me to grow in ways I didn't expect. I was able to take everything I learned and turn it into something tangible alongside an incredible team."

As the team members move forward with careers and graduate school, they're keeping in touch with dreams of advancing the project from prototype to actual circuit boards. "We were able to reach this level as students with a budget of $300," notes Alesi. As they look to the future, the possibilities are wide open.