Reading time ( words)
I recently spoke with Chico Wu of Footprintku, a company using artificial intelligence (AI) to help fill gaps in CAD library development by partnering with design teams on things like addressing resource shortages and automating custom design rulesets while also eliminating other manual steps in the process.
Nolan Johnson: Chico, can you tell us about your role and background?
Chico Wu: Thank you very much for having me today. I am the co-founder of Footprintku, with a background in sales, marketing, and customer success. We’re trying to do something a little bit different from the industry by focusing on an area of the market that has remained stagnant and been largely neglected until recently: the CAD library space. We have developed our own proprietary engine, and the technology leverages portions of AI in combination with automation and applying that in a service space. We’re a SaaS model as opposed to providing it in a utility or a tool format.
Johnson: How did Footprintku come to be, how does that SaaS model work for the customer, and where does that fit compared to other providers in the market?
Wu: We noticed that there had been a pretty significant shortage in our industry of electronics design, where finding the right candidate has been very difficult. In my prior experience at a PCB design services firm, we always had difficulty with training and retaining good librarians; we would end up needing to train them from scratch, and it was a very extensive training period—at least two years—to learn the tool, how to interpret the component datasheets, rule calculations, and the hands-on technical skills that are not in the curriculum of most graduates out of college. It’s all on-the-job training, which caused a major gap in this business because of turnover. Mr. YC Hwang, the founder of Footprintku, looked at what technologies were in the market and said, “The timing is right to leverage portions of AI and see what we can do to fill that gap for library development.”
Johnson: For the day-to-day user, how is this different? What changes for them?
Wu: We’re trying to alleviate the backlog or any queue time that an engineer or a designer has to deal with when creating their own library content. Our goal is to have someone say, “Here are my component datasheets. These are some of my criteria for building a part.” It doesn’t necessarily need to be a generic IPC; it could be a custom ruleset, and overnight or within two days, we could have that entire batch of library data built for them. In parallel, that engineer or designer can continue working on something else that’s a higher priority on their task list.
Johnson: That’s good for their time management. One business model is to have a library for parts, pick what you want, and build your library out of the components you use. Is that something that is also available for Footprintku?
Wu: That’s an adjacent market to us; it’s not the same. For example, there are a couple of established solutions in the market for that type of model. Their existing libraries are built to IPC specifications. In many cases, when you’re doing manufacturing, your contract manufacturer has a very specific set of DFA rules that are geared toward their specific equipment. What ends up happening is the user base, the engineers and designers, may retrieve some of the off-the-shelf data, but they still have to spend time modifying it to their contract manufacturer’s ruleset because it’s all about manufacturing yield and efficiency.
If they had to modify 20 or 30 parts at a time, that creates a bottleneck for other tasks that they need to execute. Our approach is to have Footprintku be familiar with the custom rules that you need, and we build your parts on-demand from the beginning before it even gets to a manufacturer for design review or to final release; imagine having that upfront during design.
Johnson: I’m getting a sense that you are looking at design teams as an ideal customer.
Wu: Absolutely. Our user base consists of hardware engineers that require schematic symbols, PCB designers that require PCB footprints, and mechanical engineers that need 3D models; thus, our deliverable set to a user would be the trifecta of all three. In some companies, those roles are gradually converging. Today, that’s something that we have seen become a higher demand, especially with the 3D model; although, in many cases, it is readily available from different internet platform sites but having it available together as part of a package with customization rules and datum orientation has been a huge benefit for our user base.
Johnson: Are simulation models a priority at all?
Wu: They definitely are. How to add more intelligence to the library data and simulation models is an area we are interested in, but for now, we have our hands full servicing the board-level platform.
Johnson: Your company name certainly is memorable.
Wu: We’re built on a premise of as much automation as possible to ensure consistency, accuracy, scalability, and turnaround. Based on that, we have our footprint robot, which is our mascot, and his name is Ku; it’s more on-demand library content.
Johnson: Obviously, you’re monetizing this. How do you engage with your customers to create a revenue stream?
Wu: Our pricing model is on a subscription basis, which we found to be the simplest for our customers as opposed to how much a component or the complexity of a component cost. We like to make it as simple as possible and allow engineers to focus more on development instead of transactional or hourly concerns. A lot of the market is either fixed or hourly, so we created a membership solution where it’s X number of parts that they would consume per month or quarter, and we deduct it off that.
Johnson: One of the statistics that has been kicking around for a while is that for the typical PCB engineers, about 30–35% of their design time is creating new library parts that they need for that design; that’s a lot of time.
Wu: And when you look at how much is involved for even a pin number mistake or other library errors, which creates a problem in functionality, and the time to market gets delayed. Then, there are also the manufacturing costs associated with redoing that. We’re trying to figure out how to support a market with as much accuracy as possible and free up engineers to do more important tasks.
Another question that comes up is, “What are we doing with AI?” AI is a buzzword, but it’s a very broad category. We look at how to mine the data from a component PDF specification sheet as well as image recognition, pattern recognition, and something as simple as differentiating between a number “1” and a lowercase letter “L” or a number “0” and a letter “O”; all of that feeds into our system and trains our robot to become more and more intelligent. We’ve tried to coin the phrase “human-robot collaboration,” or HRC, which is the combination of having an engineer or a human work in conjunction with our robot. It’s about working together to be more powerful.
Johnson: Chico, thank you very much.
Wu: You’re welcome. Thank you for having me.