From idea to product, there’s never going to be a time in the evolution of a company when it will feel as small as the early days. It’s a truly unique moment when two cofounders, perhaps with a few early hires, can sit together in a small space and riff off of each other’s thoughts, creating, developing, and shepherding a product into the world.
For Samuel Weiss, Chief Technology Officer of Instrumental, the path to this milestone has been filled with many twists and turns. Through academic pursuits at MIT and Stanford, life as a hardware engineer at Apple, and now standing at the helm of a rapidly-growing startup, Sam is using his experiences to lead a team that is building a brand new kind of manufacturing technology: manufacturing intelligence.
Sam’s path to Instrumental began on the other side of the United States, where he was finishing up his third year at MIT as a Mechanical Engineering and Mathematics dual degree. A classmate recommended that Sam apply for an internship with the company he had been working at during the past semester — Apple. Although initially hesitant (Sam was interested in pursuing research, and eventually academia), Sam decided to apply, and soon found himself in Silicon Valley for the summer.
For many, the opportunity to work at one of the most well-known companies in the world is tantalizing. For Sam, the realities of the job were long hours, cycling through projects, and a lack of ownership of the projects he contributed to. A few weeks into the internship, however, Sam started working with a new full-time hire at Apple. The semi-experienced intern and the new hire completed a few projects together over the summer, and even shared a patent (“a switch mechanism on an iPod that probably never got built,” Sam modestly interjects). That new hire was Anna-Katrina Shedletsky, Instrumental’s future cofounder and CEO.
Return to Apple
Sam didn’t see Anna for another three years. In the interim, he graduated from MIT, obtained a Master’s in Mechanical Engineering from Stanford, and began thinking in earnest about what he wanted to do moving forward. After a conversation with a former manager, Sam decided to leave academia and return to Apple, this time as a full-time employee working on a top-secret project: creating what would become known as the Apple Watch. As luck would have it, Anna was the product lead of the nascent Apple Watch team.
This time around, the experience couldn’t have been more different. “Instead of no ownership, I probably had way too much ownership,” Sam recounts. “I felt like within the first month, I was making major decisions that impacted direction of the product, if not the company.”
Sam had a major hand in creating the mechanical architecture, microphone, speaker, antenna, and the “taptic” engine for the Apple Watch; each of these technologies was novel in either its manufacturing, water resistance, or function. Within a year, Sam had become one of the leading experts in the world on the engineering challenges that went into the Apple Watch debut. The lessons learned in building a complex consumer electronics device foreshadowed many of the challenges Sam and Anna would confront head-on as they began to build a startup.
“Within a year, Sam had become one of the leading experts in the world on the engineering challenges that went into the Apple Watch debut.”
Pondering and Prototypes
After years of working together in the close-knit Apple Watch team, and with hundreds of days on assembly lines in China racked up during that time, Sam and Anna had gained two major insights:
- Trying to figure out what was happening on the assembly line by shuttling back and forth from California to China is a very slow, time-consuming process
- Many products produced in factories are still made by huge teams of people
This sparked the duo’s first idea: why not increase automation of the manufacturing process? “We thought we were going to build robots,” recalls Sam. “One of our first [half-joking] ideas for the name of our startup was Mechasaurus.”
Dreams of robots (dinosaur-shaped or otherwise), however, quickly went extinct. Anna and Sam discovered that robots were already being designed and implemented in-house by major factories in China and around the world. More importantly, they realized that robots wouldn’t provide real long-term value; however, data generated from assembly lines could. “All of the data collected on assembly lines today might tell you that you have a problem, but it doesn’t help you solve that problem,” Sam comments. “If your data says that three percent of a million units have an issue, you have to throw away thirty thousand units… but it doesn’t tell you how to fix the next million you need to build.” Currently, engineers have to look through whatever limited data they have and make educated guesses on what causes issues.
“Why is it so hard for engineers to get the data and insights they need from assembly lines?”
Another question, one that would form the basis for what Instrumental does today, began to take shape: why is it so hard for engineers to get the data and insights they need from assembly lines? From this question, Sam and Anna started laying the blueprints of a nascent Instrumental system that could help engineers solve issues with their prototypes.
Other companies soon took notice. When Instrumental signed their first client, Anna, now Sam’s cofounder and Instrumental’s CEO, was on a plane 12 hours later with a jerry-rigged prototype made from plastic containers, Legos, and a camera bought from Best Buy. The first prototype helped answer a very important development question: was data in the form of photos of units on the assembly line the right data to collect? When he was a hardware engineer, Sam often wished to easily view prototypes without having to fly back and forth for each new build, but the data collection system he and Anna had just built still needed to prove its worth for others.
After poring over images that Anna sent each day from the assembly line in China, it became clear that high-quality images were indeed a key step to showing what, if anything, went wrong in a build. As Sam recalls, they had already arrived at a critical first step, where “we had a record and we could go back and remotely analyze failed units. That was our V0.”
Over the next few months, iterations on Instrumental’s data-collection system continued. For the next version, Sam built a more robust system with a high-end camera and hired a contractor to write software to capture images and immediately upload them to Dropbox. Instead of manually sorting SD card dumps each day, photographs were now able to be accessed, searched, and sorted by serial number and timestamp. Both the growing Instrumental team and its customers could actively search records and watch builds move in the right direction.
In the summer of 2016, Instrumental deployed its own native app. Instead of logging into Dropbox, users could see, access, and zoom into high-quality images within a streamlined interface. Additionally, thanks to multiple stations on the assembly line, companies could virtually disassemble every product, allowing them to look for issues without risking damaging products.
“That was our evolution from rough prototype, to improved hardware, to something more refined that works pretty well,” Samuel reflects.
One year later, Instrumental unveiled Detect, a new machine-learning feature that enables engineers to instantly find anomalous units from their build without browsing through every unit on the computer, and certainly without having to fly a red-eye to China for an emergency inspection.
Solving Hard Challenges
“If you look at Instrumental, it’s a software company founded by two hardware engineers,” Sam remarks. “That points to the fact that it’s hardware engineers who have this problem. Mechanical engineers experience the pain of not having a useful data record. They fly to China for every build, and take apart every failed unit, but this pain is not something that mechanical engineers can solve alone.”
“We talk to engineers more senior than ourselves who have been working in the industry for 15 years and who say the technology they are using now was 15 years old when they started,” Sam continues. “It’s an underserved industry, ripe for innovation.”
Old systems don’t just hurt company engineers, of course; factory engineers and personnel are suffering from slow, antiquated systems as well. Sam likes to bring new hires to a factory so they can observe this frustration firsthand. During one trip, a QA engineer opened a panel that plugged directly into the factory, typed in commands, and sat back to wait. “It was five minutes,” explained Samuel. “But it felt like 30. It’s a horrible feeling when [the factory’s system] finally comes back and says ‘No data found’… The tools are terrible. The interface is terrible.”
Sam believes that the solution for both the companies designing products and the factories producing them starts by pairing issues in hardware with pathbreaking research in computer science. By combining his experiences as a hardware engineer with modern advances on the software side, Sam, and by extension the entire Instrumental engineering team, can design incredibly valuable tools to change how products everywhere are made.