What does it take to design an aesthetically sleek AI-driven product with mass market appeal? Three MechE MS students, Ben Wheeler (ENG ’25), Nick Teguis (ENG ’25), and Tage Colberg (ENG ’25) got into the depths of this question over the Spring 2025 term. Working with Alumni Samuel (ShengEr) Zhou (CAS & QST’ 22) and instructors Stephen Chomyszak and Eric Hazen, the team built an AI-driven candy distributing pumpkin, which earned the name “Pumpkin Steve.”
Alum Samuel Zhou owns a manufacturing firm with 8 Surface Mounting Technology (SMT) lines and has started his own firm, Shenger & AI, aiming to combine the capability of AI-driven products to mass manufacturing for the global audience. He was looking to make an AI-driven dispenser without human intervention, and reached out to Professor Chomyszak to figure out if students could make an aesthetically sleek and modern-looking prototype that would enable his vision for AI-encased consumer products. This past semester as part of their ME691/692 coursework, this team of three worked hard to realize that vision.
The team designed a pumpkin shaped candy dispenser with AI recognition that can tell when a person in a costume walks up to it, integrating the pre-trained neural network model YoloV8. This integration would not collect any data on the spot, eliminating any privacy concerns while still allowing for effective recognition, aiming to combine the capability of AI-driven products with the power of mass manufacturing to produce for a global audience.
The team overcame many roadblocks along the way. Wheeler noted, “Part of it, like facial recognition, none of us have really done much coding or electronic stuff. So, a lot of that was working with Professor Hazen and just basically learning from square one.” Colberg added, “I have learned a lot through this process, from communicating with a client to troubleshooting the prototype.” The team also had to take into consideration that this product would be mass manufactured, meaning the parts had to be easily mass produced.
Throughout the process the team was able to put what they have learned to the test as well as develop new skills. Teguis shared, “a lot of my work has been conceptual; this is the first time I get to hold something physical in my hands.” The team was surprised by how far their project has come in just one year. They are hoping to get this product into the hands of consumers for Halloween, ushering in a new age of consumer product with ingrained machine learning capabilities.
The team is now working on figuring out the next step toward turning the electronics into a PCB and working with Zhou for mass manufacturing and taking into consideration different AI-driven dispensing use cases.
Zhou encourages fellow alums to partner up with Boston University to think big and develop new consumer products in an exciting AI-driven age for variety of consumer products, his company will further collaborate with the College of Engineering in the coming semesters for more product ideations and mass production.