Manufacturing consumer electronics can be really hard — just ask Samsung, one of the biggest and best manufacturers in the world. A flaw in the Galaxy Note 8’s battery led to the smartphone overheating, which meant Samsung had to recall millions of phones.
Some estimates have suggested that the fiasco cost Samsung over $5 billion.
A new startup founded by two former Apple engineers, Instrumental, wants to make it much easier to manufacture electronics and head off complicated problems before they start costing companies thousands of dollars per minute.
Instrumental cofounders Anna Shedletsky and Sam Weiss know exactly the problems that come up when you’re manufacturing millions of complex gadgets.
“In 2014, I was at Apple leading the Apple Watch program, which hadn’t been announced yet, so it was super secret and I was working with Sam [Weiss] who was working on really cool stuff on the Taptic Engine and other components inside the watch,” Shedletsky, who is the CEO of Instrumental, told Business Insider.
It turned out that the duo had some very valuable experience. They had spent weeks on the ground in factories in China troubleshooting top-secret Apple components before they launched — experience that directly affects how they built their startup’s product, for other engineers and hardware companies that are running into the same problems as Instrumental’s founders did at Apple.
Instrumental has raised over $10 million from venture capitalists including Eclipse Ventures, First Round, and Root Ventures, including a $7.5 million Series A earlier this year.
Armed with that money, the startup is using AI and their industry experience to make it easier for any company to manufacture electronics and ship them on time.
How it works
Basically, Instrumental makes custom hardware that gets installed on a factory assembly line. Usually, these factories are run by giant contract manufacturers — companies like Foxconn that build electronics for brands like Apple and Microsoft.
Instrumental’s software allows an engineer from the design company to track specific units as they get assembled, taking photos and helping to pinpoint exactly when a problem went wrong — even if the engineer is in California and the factory is in China.
“Right now, in order to find issues, you have to physically be on the line at the right place at the right time,” Shedletsky said. “Also, when you’re in China, you’re still trying to do your job back at headquarters and keep all the balls up in the air and trying to solve the problems you’re finding.”
On an assembly line, a gadget rolls down the line on a conveyor belt, and every worker on the line has to do one thing to the device — put in a screw, or place two parts together.
Instrumental makes a hardware box that ends up right on that line, and this box takes a photo of every single individual device while it’s in production.
“So our software helps engineers do this much more quickly and efficiently,” Shedletsky said. “They can do it from their hotel after they leave the factory. They can be like, ‘What did we build today and what is useful and what wasn’t?'”
This is what the box, which Instrumental calls a station, looks like:
That data collected by the box is delivered through Instrumental’s software to decision makers at the brand that’s making the gadget.
Instrumental can pinpoint exactly when a device starts to be put together incorrectly. In the example that Shedletsky gives, she is inspecting photos of sample units of a generic smartwatch that is starting to fail water resistance tests. She flips through her software, and finds that the problematic unit had a missing screw. “The next step is, if there’s one, if there might be more. We can bring back the rest of units, and look at units before and after this one, and if see those have screws,” Shedletsky demonstrates.
Another way that Shedletsky envisions Instrumental’s software being used is for engineers to explore how the manufacturing process is going even before an issue crops up.
“Something that a lot of consumer-electronics companies care about is gaps between different parts. They care about this is because the difference between what an iPhone looks like and what a cheap Shenzhen knock-off looks like is pretty much the fit and finish,” Shedletsky said.
So using the Instrumental software, an engineer can even precisely measure the distance between two parts, or other measurements, from a photo of the device.
“Many consumer electronics companies admire that Apple quality, and so they want the same, a very tight zero-gap, or at least an even gap,” Shedletsky added.
An AI future
Instrumental’s software is going beyond just beaming photos from the factory to headquarters.
The startup was founded on the belief that increased automation will continue to replace people in factories with robots and data.
Instrumental is using machine learning to be able to spot defective units even before a human does. The startup calls this feature “detect,” and it launched on Wednesday. It processes hundreds of units and uses machine learning software to spot the outliers.
On her computer, Shedletsky pulls up a page of 50 generic smartwatches, ranked by which units are most unusual.
“We have a ‘find anomalies’ feature. It’s going through all of these images and stack ranking them in the order from the most anomalous to the least anomalous,” Shedletsky said. “It’s looking for the outliers in the whole population.”
“We use convolutional neural networks to evaluate the images,” Weiss, Instrumental’s CTO, said.
It’s these kind of machine-learning features that Instrumental’s founders see as the future of the company. In fact, when Instrumental was first founded in Shedletsky’s Silicon Valley kitchen, they wanted to build robots for factories before they decided that the software was going to be the key to their business.
“We’re building this artificial intelligence that doesn’t matter if you use our station in the end. In the end, it will be a robot with a camera on it. But our algorithms will be what’s processing that data. We’re building it to work in the factory today, but with an eye on the future,” Shedletsky said.
“We had this idea: There’s always people on the line, let’s build robots and those robots will collect data and that data will be valuable,” Shedletsky said. “A couple weeks after, we said, ‘Screw the robots, let’s focus on the data!”
Shortly after that, they realized they had a good business idea when Shedletsky was flying to China to work for their first customer. “We starting taking over all the flat spaces in her house,” Weiss said.
“So it’s your standard Silicon Valley story of just making it work out there,” Shedletsky said. “And we’ve grown the team from then. Now we’re 12 people and we’re actively growing.”