Electrical Engineering and Computer Science
Fred Buhler, Aweslome test platform

Fred Buhler Builds Better Chips for "Aweslome" Applications

ECE PhD student Fred Buhler is working on more efficient chips for a broad range of applications, including machine learning, neural networks, security, and circuits testing. A member of Prof. Michael Flynn’s lab since he was an undergrad, Buhler's work has resulted in high-profile published papers and a new company, Aweslome, LLC.

Aweslome was founded in 2016, and is currently owned and operated solely by Buhler. The company offers consulting work on custom hardware platforms, and sells equpiment designed to help small fabless and research integrated circuit (IC) designers test and characterize their own products. His products condense the functionality of bulky and expensive lab equipment into a single small platform, and show superior noise characteristics under all testing conditions.

The company is in the process of closing its first sale, and has other interested parties waiting for the next model's fabrication. Additionally, he’s already done consulting work for four companies around the state, including ECE alum Mohammad Faisal’s startup Movellus Circuits.

This platform is software-controlled and modular, allowing for flexibility that small companies and research labs typically lack. Until now, if these groups wanted a low-noise, modular, software-controlled reference, they’d look at costs of at least $5,000-10,000. Cheaper solutions mean they’d have to spend time in their lab tuning their circuits by hand, usually with a screwdriver, while interpreting the test results.

Fred Buhler test chips, old models and current
Fred Buhler holds a current model of his testing platform, with iterations dating back to his very first model laid out.

Buhler began work on this technology as an undergrad, when he spent his summer working in Prof. Michael Flynn's lab as part of the College of Engineering Summer Undergraduate Research Experience program (SURE). He had just arrived at Michigan as a transfer student the previous academic year, and wanted to get some actual research experience.

The early models of his device saved Prof. Flynn’s research group $100K in test equipment while simplifying the entire testing process. Now, several redesigns and a business later, the group is purchasing Buhler’s latest model for the lab as one of his first customers. He was helped along the way by courses and a grant from the Center for Entrepreneurship, funding the assembly of his most recent model.

Fred Buhler reserch neural network chips
A demo printed circuit board for Buhler's second project, a neural network using analog computation.

Buhler remained with Prof. Flynn’s lab when he started his PhD studies, after graduating summa cum laude in electrical engineering from Michigan. He now works on a variety of analog-digital circuits, focusing in his first two years on using analog circuit blocks to do efficient computation for machine learning and neural network applications. His research has already resulted in two articles published in the IEEE Symposium on VLSI Circuits.

Incorporating analog computation into a digital system gives the benefit of faster, more efficient computation for many of the core processes that drive machine learning. This comes without designers needing to abandon their usual digital work interface.

Currently, Buhler is working on a new chip for nuclear security applications. This chip will address the problems that arise when two nations need to measure each other’s nuclear arsenals, without revealing classified information.

For the moment, Buhler is content to keep Aweslome small and manageable. With products that appeal to a niche of research startups and university labs, he doesn’t expect it to grow into a giant – but he’s already received interest from a number of labs around the country.

“I’ve had at least three visiting researchers tell me they’d be my first customer,” Buhler says, referring to guest lecturers who have toured the lab. “I’ve got plenty of other opportunities if this doesn’t grow into something that could sustain me on its own,” he says. “That takes the stress off, and makes this whole process very enjoyable.”

Fred Buhler

Research Interests

Using analog circuits for efficient computation in machine learning, neural network, and security applications.

Aweslome, LLC
Fred Buhler Aweslome Logo

Design, troubleshoot, and characterize hardware platforms with a focus on designing and interfacing custom hardware with embedded software. Products that address pain points for small fabless and research IC designers in testing and characterizing ICs, specifically software programmable ultra-low noise voltage & current references.

A Non-Traditional Student

Before starting college, Buhler spent eight years in the Navy, deployed to a ship in Japan. He worked on electronics engineering, and eventually got to teach in a base south of Tokyo. He wasn’t a great student in high school, so coming directly to Michigan wasn’t an option. But after excelling at a Community College, he transferred to Michigan to pursue his interests in electrical engineering.

“I try to make the most of all that’s offered at Michigan,” says Buhler, who’s seen how much narrower one’s options are without the opportunity of a great education.

Buhler enjoys his experiences with other non-traditional students in Prof. Flynn’s group (Flynn enjoys working with individuals of diverse backgrounds). His first mentor, Jeffery Fredenburg, for example, was an electrician for eight years before he started his PhD.

Posted July 26, 2017