The technical documentation for chips in the semiconductor industry is often thousands of pages long. Keeping those documents updated with every revision is a massive lift, as is generating the manuals and tutorials for engineers who then implement those chips in their own products. And to make matters even more complicated, very few products only use a single chip, even if they are seemingly as straightforward as a power tool.
Driver, a startup coming out of stealth today and announcing an $8 million seed funding round led by GV, uses AI to cut this entire process down to only a few hours, while also giving businesses the opportunity to generate user-specific documentation.
The company’s CEO and co-founder Adam Tilton launched (and sold) a number of startups, including wearables startup Rithmio, before landing at Nike after it acquired his edge computing and machine learning startup Aktive. While at Nike, he met his co-founder and CTO Daniel Hensley, who at the time was part of the leadership team of a small firm that helped clients integrate embedded devices and use machine learning. After leaving Nike, Tilton went to work at healthcare startup Levels, where he led signal processing for a hardware project. To round out the team, they also brought on Jimmy Hugill as co-founder and CFO.
“Throughout my career, I’ve done a lot of development for embedded technologies that were to then be used by others,” Tilton said. “At Rithmio, we had a library of functionality that wearable product manufacturers could use to enable some functionality. At Nike, I was a consumer. I would buy components. I was on the selection committee to take the components — and then we were trying to build a product out of them. And I have just over and over in my career, had to work on this particular challenge.”Image Credits:Driver
The documentation for these components is often very low level, with example code that can be out of date and focused on only a single programming language.
“Daniel [Hensley] and I were doing a lit review to get deep into the technology and I had the idea that I should go into the PDF and pull out each of the different APIs, and go into the example code, and pull out the example code, and effectively get ChatGPT to do the thinking about how I’m supposed to do this in Python, because I wanted to wrap all of this into a Python program integrated into our automated testing system,” Tilton said.
After seeing this, Hensley suggested building a software platform to help engineers do this at scale — one that caters to both the suppliers of chipsets like microcontrollers, ASICs (application-specific integrated circuits), and FPGAs (field-programmable gate arrays), as well as the engineers who then need to translate those documents for their own products.
Even within a company, much of the internal documentation for building a certain product can become outdated as new components are introduced. Driver, which is language-agnostic, aims to be a solution for those use cases, as well as helping buyers and sellers keep their internal documentation up to date.
The company promises that it can speed up a team’s understanding of a new codebase by 50% and save companies weeks of employee time that would otherwise be spent writing documentation.
The company’s seed funding round was led by GV, with participation from Y Combinator and “over a dozen early-stage and angel investors.”
“We invested in Driver early given our excitement about the founders and the novel use cases of generative AI they are addressing with a large untapped market opportunity,” said Luna Schmid, partner at GV. “The founders have an incredible track record of building and bring extensive experience and hard-earned lessons from working with complex codebases. We believe Driver is a game changer for any team that needs to document complicated technology quickly and ensure it can be understood by all constituents.”