In twelve weeks you'll have built a robot that drives itself.
Not watched one. Built one — a machine that maps a room it has never seen, works out where it is inside that map, and drives itself somewhere without you touching it. It'll be on your GitHub with the code, the write-up and the video, and you'll be able to explain every line of why it works. That artifact is the thing standing between you and a robotics job right now, and you can start it twenty minutes from now on the laptop you're reading this on.
You've done the tutorials. You still can't build a robot.
You got a talker and a listener printing to a terminal. You followed the six-part series. And then the trail went cold, because the next step quietly assumed a lab, a mentor, and four years you don't have. Somewhere in there you started to suspect the problem was you.
It isn't. You already ship distributed systems that are harder than this. What you ran into is that robotics material is ordered backwards — it hands you the framework first and hopes you'll infer the problem it solves. You can't learn that way, and neither could I.
So here you write the robot first, in a language you already speak, and watch it fail in specific ways. Only then do you meet the professional tools — at the exact moment they stop looking like ceremony and start looking like the answer to a problem you personally just hit.
Three levels
Pick the one that matches where you are today.
Each level runs a fixed twelve weeks and is its own decision. No auto-renewing subscription, no surprise charge on day 84. Read them as three different finishing lines and choose the one you actually want to be standing on.
Level 1 — start here
Roboticist Foundations
Twelve weeks from nowYou have an autonomous robot in simulation that builds a map of a world it's never seen, localizes itself in it, and navigates it — written by you, on your GitHub, with a demo video. You can read a robot's failure and name the cause.
What that opensYou stop being someone who's "interested in robotics" and become someone with evidence. That's the difference between an application that gets ignored and one that gets a first call.
Twelve weeks from nowThere's a robot you built with your own hands driving around your home, finding its way to places you tell it to go, and doing a job you chose for it. You know why every wire is where it is.
What that opensThree separate portfolio pieces — control, navigation, perception — each aimed at a different kind of robotics team, each backed by numbers you measured. You can sit in an interview and defend all three.
Twelve weeks from nowYou've done original work on a problem you chose, made it work in the physical world, and written it up to conference standard with your name on it. Three people at a time, so it's genuinely your problem we work on.
What that opensA paper is the currency of graduate admissions and research teams. It's also the clearest possible proof to an employer that you can be handed something nobody has solved yet.
You can't just buy the top two levels. Both depend on me knowing where you're starting from. Put someone in the wrong program at the wrong time and they end up with a half-finished robot and a worse opinion of robotics than when they arrived — I've watched it happen, and I'd rather it didn't happen here.
So we talk first. You tell me where you are and what you're trying to build; if I think we can get you there, I'll send you the numbers. If I don't, I'll tell you what to do instead.
What opens up
Robotics stopped being a niche. It became infrastructure.
2014
221,000
2017
400,000
2020
390,000
2023
541,000
2025 — record
621,000
2028 — forecast
700,000+
New industrial robots installed worldwide, per year. 2014–2023 from Our World in Data, compiled from IFR World Robotics. 2025 preliminary figure and 2028 forecast from the International Federation of Robotics, reported June 2026.
More than half a million new machines go in every year, on top of an installed base that has passed 4.6 million. Nothing at that scale stays a specialism — it becomes infrastructure, and infrastructure has to be programmed, integrated, tuned, monitored and repaired for every year it keeps running.
That's the part people miss when they picture a robotics career. The work isn't mostly designing new robots. It's the much larger job of making fleets of them do something useful and keep doing it — and nearly all of that is software, written by people who understand what the machine is physically doing.
You cannot bluff that in an interview. Either you have made something move in the physical world or you have not, and AI hasn't changed it. A model will write you a flawless Kalman filter, and you will still be the one at midnight working out why yours drifts after two minutes, because that answer is on your floor, not in any training set.
Reported pay at US robotics companies runs from $100k to $245k.
Compensation: self-reported medians for software engineers at Serve, Nimble, Path, Rapid, Open Robotics and others via Levels.fyi, retrieved August 2026. Small samples, skewed toward well-funded firms in expensive cities — it shows you the ceiling, not your offer.
What you get from me
Someone who has already made these mistakes, so you don't spend a month on each.
You lose weeks, not months
The three weeks you'd burn on the wrong approach get cut to an afternoon, because I've been down that road on machines deployed for NASA, Yamaha and Boeing programs — the kind you can't SSH into after launch.
You learn what's current
I still build and deploy systems as a project scientist at Carnegie Mellon. What you get is how a problem is handled this year, not the year I stopped practising.
You can debug it alone afterwards
Anyone can hand you a robot that works. The skill worth paying for is the twenty minutes after it stops working — and you keep that skill long after the twelve weeks end.
He watched all the videos too. Here's what actually changed.
"There are tons of videos on the internet and everything else — people show videos where you connect, type, and it works — but I couldn't get started. Even listening or watching, I just couldn't begin. That was my biggest challenge."
"The transformation has been real. If you threw me code before, I wouldn't engage with it after a certain point. But now I actually try to read it better. That's the difference I see. I'm not 100% there yet, but I'm confident I can use the web and try to build upon things. It just opened my mind."
Supply chain professional, moving into roboticsTwo months of one-on-one coaching with me
That shift — from bouncing off the material to being able to pick it up and go — is the thing I'm actually selling. The robot at the end is how you prove it happened.
Students and parents
Still in school? You'd finish this ahead of most undergraduates.
Most of what's in Foundations isn't taught until second or third year of an engineering degree. Finish it in high school and you arrive at university already able to do the thing your classmates are still reading about — and you'll spend your first year building instead of catching up.
On a college application it's the difference between "interested in robotics" and "here is a thing I built, here is the code, here is why it works." Admissions officers see a great many of the first kind.
It runs entirely on a laptop, so there's nothing to supervise: no soldering iron, no batteries, nothing that can burn a desk.
I write one long essay a week about what actually happens when robots meet the physical world — the failures, the field notes, the things that only show up at three in the morning in a car park. It's free, and it's the best way to find out whether you can stand listening to me for twelve weeks.
I also post regularly on LinkedIn, X and YouTube. Same material, shorter form.