What you get
A publication-quality technical paper, written to conference standard, on work you did — with my guidance and co-authorship. Realistically an arXiv preprint inside the twelve weeks. You own it and can submit it anywhere.
Level 3 / one-on-one / 3 slots
You'll have done work nobody had done before, and written it down.
Twelve weeks, just the two of us, on a problem you bring. At the end there's a robot that works in the real world and a publication-quality technical paper with both our names on it. If you're applying to a top robotics program, that paper is worth more than everything else here combined — almost no applicant from industry has one.
Specification
Cohort1
Term12 weeks
Your effort15–20 hrs/wk
CallsWeekly minimum
Slots3 at a time
OutputRobot + paper
AdmissionInterview + scoping call
The paper
What I will and won't promise.
What I can't guarantee
Acceptance at ICRA, IROS or RSS. Those deadlines are annual and fixed, and acceptance depends on novelty and results that neither of us controls yet. Anyone promising you a published paper in twelve weeks is lying to you.
What opens up
Three doors that a paper opens and a portfolio doesn't.
Graduate admissions read differently
A PhD application from industry usually has no research output in it. Yours would have a paper and a supervisor who can write specifically about how you handle an open problem — which is the exact thing admissions committees are trying to guess at.
Research and advanced-team roles
Teams doing new work hire for evidence you can operate without a spec. Published work is the cleanest form of that evidence, and it moves you out of the pile of candidates who have only ever implemented someone else's design.
You stop needing me
The real outcome. By the end you've chosen an approach, been wrong, recovered, and shipped — which means the next hard problem is one you can take on alone. That's the whole point of the twelve weeks.
How it works
It starts with your problem, not my syllabus.
Week zero — scoping
We define what you're building and what "working" means for it. I write you a twelve-week plan with hardware choices, a bill of materials, the electives we'll draw on, and the failure modes I expect you'll hit.
Weeks 1–12 — building
An hour-long call every week, minimum, and more when you're blocked — a stuck week is wasted money for both of us. Between calls I'm in your repo and your messages.
What you're actually buying
Judgment, not hours. Choosing your hardware, reviewing your architecture, debugging on a shared screen, and above all stopping you from spending three weeks on the wrong approach. You'll do the work — 15 to 20 hours a week of it. My job is making sure it's the right work.
What you end with
A robot that does the thing, a paper, and the judgment to build the next one without me — which is the point.
Electives
The menu we build your plan from.
We'll pick two or three of these on the scoping call, chosen against what you're building and where you're trying to end up.
Controls & estimation
Balancing robots, inverted pendulum, LQR, MPC, custom EKF and UKF, visual-inertial odometry.
Aerial
Drone dynamics, flight control, state estimation under fast dynamics.
Manipulation
Arm build, MoveIt, inverse kinematics, grasping.
Outdoor & field
Solar power budgeting, GPS, ruggedization, weather, remote operations. This is what I spent fifteen years doing and nobody else teaches it.
Custom hardware
PCB design and fabrication, CAN bus, harnessing, bringing up a board you designed.
Learned control
Reinforcement learning, sim-to-real transfer, Isaac.
Fit
Three slots at a time, and I mean it.
This is the only thing I sell where I'm the bottleneck. If all three are taken I'll tell you when one opens rather than take your money and thin myself out. Forge graduates get first call on an opening.
One more thing, said plainly: if you can't already program professionally, this isn't the right door — and I'll tell you that on the call rather than after you've paid. There's a different route for you, and we'll talk about it.