Build a Humanoid Robot From Scratch.
Master ME, EE, and AI.
An elite 22-week cohort for high school students to construct, simulate, and train ToddlerBot — the open-source, ML-compatible humanoid platform. Mentored by Genboo engineers with direct technical consultation from the paper's first two co-authors.
Aiming at top engineering programs. ToddlerBot becomes a portfolio cornerstone — and a story admissions officers haven't heard before.
Already shipping projects solo. Want mentorship from working researchers and access to a real lab setup instead of guessing alone.
This is what your robot will do.
Whole-body coordination. Balanced contact forces. A learned policy transferring cleanly from simulation to the real chassis — pushing a loaded cart across the lab floor. That's the end state we engineer toward together.
- Train an RL walking policy in MJX, then deploy it onto the robot you built.
- Calibrate contact dynamics so the robot can apply real force to the world.
- Ship a portfolio-grade sim-to-real demo by the end of the cohort.
A research-grade humanoid.
ToddlerBot is the Stanford-published, open-source humanoid platform you'll build, wire, and teach to walk. Every component is portfolio-grade.
30 Active Degrees of Freedom
7 DoF per arm, 6 per leg, 2 in the neck, 2 in the waist. Ultra-robust 3D-printed structural components engineered for repeated drops, falls, and self-recovery.
- Anthropomorphic 30-DoF kinematics
- Dynamixel-class actuator integration
- Print-in-place tolerancing & assembly
Jetson Orin NX Onboard Compute
Stereo fisheye perception, custom power management, and a clean serial bus topology — production-grade electrical design, not a hobby breadboard.
- Stereo fisheye camera pipeline
- Custom DC power distribution
- Smart-bus actuator daisy-chain
Sim-to-Real Reinforcement Learning
Build a faithful MuJoCo / MJX digital twin, identify system dynamics, and train RL walking policies that transfer to the real chassis.
- MuJoCo + MJX physics simulation
- System identification (sysID)
- RL omnidirectional locomotion
Three phases. One robot. Six hours a week.
Summer Intensive — The Physical Build
Assemble the complete 30-DoF chassis, wire every electronic subsystem, calibrate joint zero-points, and deploy Python keyframe scripts so the robot performs open-loop push-ups and squats.
In-person lab build sessions on weekends during summer, plus weeknight virtual office hours throughout the fall academy.
Fall Academy — The Digital Brain
Stand up a MuJoCo digital twin, perform motor system identification, train RL walking policies in MJX, and explore real-time VR teleoperation or vision-based manipulation.
Pick a research direction with your mentor. Here's what past and current cohorts are chasing — your capstone can match one or chart its own:
- Whole-body manipulation — pushing a loaded cart across the lab floor (shown in demo)
- Loco-manipulation — opening doors and drawers mid-stride
- VR teleoperation handoff — a human pilots, the robot reproduces in real time
- Vision-conditioned grasping — finding and picking objects on a cluttered surface
- Multi-robot coordination — two ToddlerBots hand off a payload
- Sim-to-real on novel terrain — slope, gravel, or compliant surfaces
Weeknight virtual seminars and code reviews, plus monthly in-person lab integration days for sim-to-real testing.
Output Sprint — Cohort Showcase
Turn your capstone into shareable artifacts. Write up your work in research-paper format, ship an open-source contribution back to the ToddlerBot ecosystem, and present an 8-minute talk at the cohort showcase — same room, same standards as a top-lab progress review.
Most programs end at "I built it." This phase makes you fluent in the language top labs and grad-school admissions actually read: a paper, a working repo, and a talk. You leave with an artifact you can hand to any researcher.
- Reviewed line-by-line by your mentor before submission
- Optional provisional patent draft for novel methods
- Talk recorded for your application portfolio
- Alumni intros to humanoid robotics labs
One in-person showcase day for talks & demos, plus daily virtual writing rooms and review sessions throughout the sprint.
22 weeks, mapped to real engineering deliverables.
Each module is taught lab-first: a short concept primer, then hands-on time on the bench or in simulation, then a code review with your mentor.
You're not following a tutorial. You're shipping research.
Every cohort student builds against the same platform Stanford's research team publishes on. Your code, sims, and hardware get reviewed by the people who wrote the paper.
The first two co-authors of the original ToddlerBot paper serve as Technical Consultants — reviewing student code, simulation designs, and hardware builds.
Working roboticists run weekly labs, code reviews, and 1:1s. A strict 1 mentor : 4 students ratio so your mentor knows your code, your robot's quirks, and your goals — not just your name.
Your improvements to drivers, sim tooling, and policies can land upstream in the ToddlerBot repository — a portfolio piece that compounds.
Stanford-published open-source humanoid research platform.
Seats are strictly limited.
Due to hardware complexity and lab space constraints, this cohort is highly selective. Ideal candidates have prior exposure to programming (Python / C++) or competitive robotics (VEX / FRC) — but passion and problem-solving drive are prioritized over credentials.
Selected applicants discuss tuition, hardware kit costs, and scholarship eligibility once shortlisted. We're committed to making the cohort financially feasible for students who belong here.
Application received
Thanks, future roboticist. We'll reach out within 5 business days.