Embracing AI-based electronics design: a conversation with CELUS

For these engineers, bogged down by thousands of different data sheets, formatted differently from company to company, and facing the common issue of component obsolescence, AI-based electronic design is emerging to solve these challenges and support more streamlined design.

Tobias Pohl, Co-Founder of CELUS, is doing exactly that, and spoke to Electronic Specifier about the impact he witnessed AI-based design having on the electronics industry. 

CELUS was founded 7 years ago, out of the “engineering challenge”, as Pohl saw it. These challenges are multi-faceted, some of which the company has solved, he said. 

Arguably one of the most frustrating challenges was component selection: “I went through the entire design process … I picked my components … afterwards find out that two of them are currently not available. So I needed to go back and find out [how] to replace them … It was shocking to me how often that happened,” explained Pohl.

Rob Telson, VP of Sales at the company, agreed: “What happens is, for example, if you choose a host CPU or processor to build a device, you have to have the appropriate interfaces and surrounding products that work with that device, and then [if] you find out that one of those interfaces isn’t available, then you have to go back, change the CPU and start the whole process all over again.”

This commonly shared experience of the design process breaking down due to components becoming obsolete was a core part of why CELUS was founded. More on how their platform works later.

The huge number of component options engineers can design with can be overwhelming, and they can return to components they’ve already used and know works well. This means manufacturers pushing their newest component may lose out because the engineer won’t necessarily devote a huge amount of time to finding the latest microprocessor, for example.  

“Manufacturers trying to push everyone [onto] the newest set of components is … crucial, because they ultimately want everyone to switch to the newest generation, so they can phase out the old generations, and not keep massive stocks or multiple … lines open forever,” said Pohl. 

Manufacturers ultimately phase out old components, triggering a forced redesign for engineers who are creating products with life cycles of 10-15 years and realise they need to find something else.

“With solutions like ours … on where you dynamically find the best fit for every design you build, that problem … kind of disappears because you don’t have that time restriction anymore,” added Pohl. “We’ve seen plenty of examples of entire production lines … stopping because of one chip being missing.” 

CELUS’ platform

Enter CELUS’ platform. It is designed to be an intuitive experience for engineers while they are designing products, offering features such as generating whiteboard sketches on the platform and searching for available components to make it easier. 

“We check what components are out there, what are compatible with one another, what schematic you need to use [for] that specific component,” Pohl outlined.

It can provide the engineer with recommendations, at which point they are able to confirm these are the components they want, or keep searching. The AI functionality is used to sift through huge amounts of data and present it in a digestible way.

“For context, there are currently about 600 million components out there. So if I, for example, open a data sheet for a Renesas chip, I’ll probably be looking at 100 pages … that I would need to go over to make a deep decision,” explained Pohl.

Once all the main components of a design have been confirmed by the engineer, CELUS can “take it one step further” by suggesting underlying components like the final resistor needed for a board, as well as provide lead time estimations, life cycle status, unit prices, and so on.

The data the company draws on to enable easy searching of components includes pure component data - what is displayed in a bill of material, for example - and digital twins, AKA ‘cubos’. 

“They are like the digital twin of an IC [and] is what we build in collaboration with the semiconductor companies,” said Pohl. “For example, Renesas, one of our customers that we work together with, gets direct data feeds and, with that, integrate their solutions onto our platform.” 

What does AI-based design look like?

The platform used machine learning models in its early days, “before it even ever worked”.

“It took us roughly three years to reach a point where you could hit a button and the schematic that’s correct, works,” said Pohl. “There are other parts … that back then was just … not feasible.” 

Its early embrace of AI seems poignant now, where companies are exploring how they can integrate AI to improve efficiency, save time, and reduce costs.

So, what does AI-based electronic design look like?

Pohl mentioned two specific scenarios where AI performs well: data handling and recognising that people have different ways of designing - no one person approaches it the same.

“Data sheets are structured differently because they’re not [from] one manufacturer … they don’t structure data the same way … this is where AI can make a big difference … AI models can take that messy data and [turn it] into something machine readable.

“It’s not just the data that is messy. Human thinking or the way we structure a design, the way we think about a design [is different] … now we can say, tell us however you want to say it, in your language, in your phraseology … and we translate that into the standardised data set that we have.” 

Welcoming AI-based electronic design

Pohl hopes that engineers will welcome AI-based electronic design to resolve common issues they face. Its ability to streamline design, reduce time and identify the best components for the engineer is a winning combination.

For the component manufacturer, guiding engineers to the right components is hugely beneficial. A disparity in time and resources means it may be easier for a component manufacturer to deal with a large-scale account, like an Apple, who will have multiple different engineers working on a product as opposed to one or two at a small company. This is in spite of the “majority of revenue” that comes from small companies, Pohl mentioned.

In relating electronics design to how software development used to operate – i.e., there being capable software engineers but few apps and complex systems that were difficult to access - Pohl said he hoped that the same transformation could happen to the electronics industry.

“The same engineers with the same creativity can build so many more solutions … the creativity of an engineer is what limits how many designs we can have,” he said.

“This is a revolution going on,” said Telson, “especially with the younger engineers … they’re very accustomed to looking for intelligent solutions and to get them started.

“We know that because when we launched it [the platform], our goal was to have 7,000 users on the platform. In a year, we had 21,000 users.”