IPC International Community magazine an association member publication
Issue link: https://iconnect007.uberflip.com/i/1543955
14 I-CONNECT007 MAGAZINE I MARCH 2026 I spent 16 years at Altium, where I gained expo- sure to different types of electronics design and industry verticals. I was an FAE, a product man- ager, and an electronics and PCB subject matter expert. In 2020, during the pandemic, I moved to Autodesk as a senior product manager for the Fusion 360 Electronics product. I was involved in significant advancements, including a partnership between Autodesk and Ansys to integrate a field solver and provide parameter extraction and signal integrity capabilities. Then I was a freelance engi- neer until I came to Quilter. In your work as a PCB designer, what made an impression when it came to the concept of auto- routing and any sort of AI assistance in design? Throughout my career, I have had ongoing con- versations with experienced designers about automation, particularly auto-routing. Some of the results early on were admittedly terrible. But why did designers recoil so strongly at the word "auto-router?" I think part of it was aesthetics, that auto-routed boards just didn't look like a human designer had done them. But mostly, earlier auto- routers produced poor results that required exten- sive setup and manual cleanup. In my research, I learned about the graph-based algorithms used to develop the auto-routing technology. For instance, "NeuroRoute," which Protel (later renamed to Altium) acquired in the 1990s was based on neural nets and topology. That early approach was sound but limited by available training datasets, the very limited computing power at that time, and required significant user manipulation. I wrote about it in an article in 2018, and in 2022, Quilter founder Sergiy Nesterenko messaged me on LinkedIn—which I unfortunately missed at the time. However, in March 2024, as I was research- ing AI startups, I began reaching out to founders, and I came across Quilter. I also found some podcasts with Sergiy, who was very articulate and shared his vision superbly. That's when I saw, sure enough, that he had mes- saged me on LinkedIn, complimented my 2018 ar- ticle, and said he'd like to connect. Quilter already had a functioning pre-alpha prototype that any- body could try, and examples of boards that they had built. They had made a small camera board with an STM32 microcontroller that could perform basic face detection. I thought that if he had gotten that far in such a short amount of time, he would have a real shot at solving the auto-route problem properly and unde- niably, even if it doesn't look like a human did the work. To be honest, I don't care if my trace routings are 45 and 90 degrees. I will design at any angle, so auto-routing is fine. I care more about signal integrity and functionality. What were you looking to do at Quilter? My goal was still to shift the industry perspective from fearing job replacement by AI to embracing better tools. There are not enough board design- ers; every company I talk to struggles to get their boards through the pipeline. For any single elec- tronics product in mass production, anywhere from 10 to 100 boards are designed during R&D. Most iterations never leave the building, so the indus- try is itching for design automation tools that will remove some of the "busy work" and allow skilled designers to solve the more difficult problems. Figure 1: NeuroRoute User Manual front cover. (Image provided by B. Jordan)