Robotics, control, machine learning, and agentic software. Short writeups on what I built, why, and what I learned.
Internal CI reviewer that fans out across multiple agent lanes, then gates every finding through a verification chain before it posts.
Joint project on running NMPC trajectory tracking on a real quadcopter. I own the airframe and the PX4 integration; my teammate leads the controller.
A neural network learning to imitate an NMPC path follower on a Pi-based office robot.
Trained an RL policy to drive a 1/8-scale Ackermann car around a taped track, then deployed it on real hardware. Training the policy was hard; so was crossing the sim-to-real gap.
A small instrument panel of what I'm watching, building, and learning this month.
I am Josh Kannemeyer, a robotics, machine learning, and agentic-software engineer based in Essen, Germany. I learn by building things and pulling them apart: pick something I don't understand, build the smallest version of it that runs, break it on purpose, and chase the failure until the system makes sense. I have an MSc in Automation and Robotics from TU Dortmund. This site is where I keep project notes, write through ideas, and park the small demos that make technical tradeoffs easier to see.
download CV
