October 27, 2018
I started a new job at Starsky Robotics yesterday. Their people ops manager asked me to write an answer to the question, "Briefly describe your past work or school experience that led you to this work".
I was hand-etching printed circuit boards and working on cars since before I had a driver's license. I was pretty sure I'd end up doing hardware after college (I have several close friends who went into chip design) but I met Bill Gates in 2005 and he convinced me the real action was in software. Software also seemed to be where most of the smart people I knew ended up. I spent the next ten years building just about every kind of software system—a database engine, a hotel booking system that moves millions of dollars, an email productivity tool, analytics that tracked app performance across billions of mobile app installs, a natural language-based booking system, rental property management software, a music-based ad network, a project dashboard system that generated Powerpoint slides automatically, even an e-commerce paint site that matched colors based on metamerism—but never really found a "specialty".
When I wrapped up my last thing, I decided I wanted to go really deep in something, so I looked for what I could see myself doing for the next 10 years or longer. It had to be something that was going to be relevant on that time scale, something with a lot of technical depth, where I could apply my breadth of knowledge across hardware, probability/stats, nuts-and-bolts software engineering, and something with a lot of business impact. I settled on fixing healthcare or autonomous robots and ultimately chose robots because I think the challenges with healthcare are more regulatory/political, but we're on the cusp of a massive golden age of innovation in autonomous robots.
So that's what I'm doing now.
The place is better than it looks. They're a little under-the-radar—no mass-market brand like Coinbase, Stripe, or some of the other places I considered—but they're a small team that's been able to do a lot.
These guys and gals have real breadth; autonomous vehicle control means executing across (1) hardware design (lots of cameras, sensors, and ruggedized electrical design), (2) perception (making sense of all the data from the sensors using machine learning and lots of math), (3) motion planning (both local and global), and (4) cloud services/ops.
One thing I noticed already: Starsky has a very different engineering culture than I'm used to. Most places I've worked have had teams of people who, while not exactly interchangeable, could at least "pinch hit" for each other, e.g. an infrastructure developer might be able to throw together a crappy web UI or mobile app. At Starsky, if the cloud ops people wanted to work on the low-level C code that turns the truck's steering wheel, they wouldn't have a clue how to do it. That's new.