PROGRAMME · TECH

On request

Engineering Course: 12 weeks to your own Physical AI stack.

For engineering teams who want to bring humanoid applications into production themselves. Week 1 is a defined kickoff (hardware, ROS 2, VR-teleop, Isaac Sim, AI for Robotics). Weeks 2-12: every week a half-day on site, under guidance, building on your own humanoid. Own studio assigned, quarterly cohort, limited seats.

Format

12 weeks · kickoff week + 11× half-day per week on site + homework

For whom

For engineering teams (4–8 people).

Format

Engineering team, 4-8 people

Pricing

Bespoke

12-week Engineering Course

Price depends on team size, use case and hardware choice

On request

What you do

What happens during the quarter?

The programme starts with a five-day kickoff week in which your team touches the whole Physical AI stack hands-on, from safety and ROS 2 to VR-teleop, simulation in Isaac Sim, and imitation/reinforcement learning. Friday ends with a working mini-use-case on your own work shop floor. Then for eleven weeks you work half a day per week on site on your own humanoid, under guidance from MICS engineers. End point: a pilot in production plus the skills to keep building independently.

Next cohort: starts September 2026. One cohort per quarter, limited seats.

What you take away

A working pilot for your use case, deployed in production
Substantiated business case and roadmap for scale-up
A team that can build, deploy and maintain humanoid applications independently
Mentorship relationship with MICS engineers during and after the programme

WEEK 1 · KICKOFF

Five days of Robotics fundamentals.

The kickoff week is the defined starting week of the programme. Every morning a learning block, every afternoon hands-on, every day ends with a concrete exercise. Friday ends with your own mini-use-case on your work shop floor.

Day	Learning block (morning)	Hands-on (afternoon)	Exercise
Day 1 · Fundamentals & Safety	Safety, e-stop, hardware overview, Jetson/Linux setup, ROS 2 fundamentals (nodes, topics, DDS)	First robot startup, e-stop drill, first publisher/subscriber, visualise state in Rviz	Robot live in protected mode + state stream in Rviz
Day 2 · Stack & Teleoperation	Sensor stack (depth, LiDAR, IMU), robot state estimation, VR-teleop architecture	VR-teleop with Quest 3, dataset recording via LeRobot	Record 10 minutes of teleop data for a pick-and-place demo
Day 3 · Navigation & Simulation	Isaac Sim + MuJoCo intro, collision avoidance, sim-to-real principles	Import robot into Isaac Sim, waypoint navigation, first sim run	Identical run sim vs. real robot, name the gaps
Day 4 · AI for Robotics	Imitation learning (behavior cloning), reinforcement learning basics, data curation	Train policy on data from day 2, evaluate in sim	Deploy trained policy on the real robot and debug
Day 5 · Applications & final assignment	End-to-end pipeline, arm + whole-body control, deployment & MLOps for robotics	Team works on its own mini-use-case on your work shop floor	Demo + retrospective + roadmap session

Day 1 · Fundamentals & Safety

Learning block (morning): Safety, e-stop, hardware overview, Jetson/Linux setup, ROS 2 fundamentals (nodes, topics, DDS)
Hands-on (afternoon): First robot startup, e-stop drill, first publisher/subscriber, visualise state in Rviz
Exercise: Robot live in protected mode + state stream in Rviz

Day 2 · Stack & Teleoperation

Learning block (morning): Sensor stack (depth, LiDAR, IMU), robot state estimation, VR-teleop architecture
Hands-on (afternoon): VR-teleop with Quest 3, dataset recording via LeRobot
Exercise: Record 10 minutes of teleop data for a pick-and-place demo

Day 3 · Navigation & Simulation

Learning block (morning): Isaac Sim + MuJoCo intro, collision avoidance, sim-to-real principles
Hands-on (afternoon): Import robot into Isaac Sim, waypoint navigation, first sim run
Exercise: Identical run sim vs. real robot, name the gaps

Day 4 · AI for Robotics

Learning block (morning): Imitation learning (behavior cloning), reinforcement learning basics, data curation
Hands-on (afternoon): Train policy on data from day 2, evaluate in sim
Exercise: Deploy trained policy on the real robot and debug

Day 5 · Applications & final assignment

Learning block (morning): End-to-end pipeline, arm + whole-body control, deployment & MLOps for robotics
Hands-on (afternoon): Team works on its own mini-use-case on your work shop floor
Exercise: Demo + retrospective + roadmap session

WHAT YOU GET YOUR HANDS ON

Unitree G1 EDU: humanoid robot on the work shop floor.

Every engineering cohort gets its own G1 EDU assigned for the quarter. Open SDK, modular, and dexterous enough for real-world use cases.

Height / weight: 132 cm · ~35 kg
Degrees of freedom: 23-43 DoF (config-dependent)
Sensors: Depth camera + 3D LiDAR + 4-mic array
Compute: 8-core CPU + optional NVIDIA Jetson Orin
Hand option: Dex3-1 3-finger hand · 7 DoF
Runtime: ~2 hours per battery swap (9000 mAh, quick-release)

WHERE YOU'RE HEADED

Levels of Autonomy: robots growing into coworkers.

The 12-week programme moves your organisation from Level 1 (tele-operated) toward Level 3 (classical autonomous), with the skills to keep scaling to Level 4 and 5 in the years after.

Level 1 · Remote-Controlled

Operator drives the robot directly; robot makes no decisions.

Safe reach into tight or hazardous spaces without sending people in.

Level 1a · Tele-operated motion suit

Operator tele-operates via VR suit; robot learns while doing.

Hybrid labour from a different geography, or training-dataset build-up.

Level 2 · Scripted

Robot repeats what it was taught; operator starts/stops and intervenes.

Automate very repetitive subtasks without major line changes.

Level 3 · Classical Autonomous

Robot follows rules and local plans; operator supervises.

One robot serves multiple similar stations, reducing bottlenecks.

Level 4 · Skill-Level AI

Operator gives high-level instructions; robot chooses how within a task family.

Fast retasking across variants with minimal engineering effort.

Level 5 · Agentic Fleet Co-Worker

Humans set goals and constraints; fleet plans work and escalates only when needed.

Humanoid workforce layer that optimises plant-wide throughput and quality.

Enrol

Individual or whole-team enrolment, both work. Leave your details and Bart will schedule a short intake and confirm your spot in the next cohort.