Engineers and project teams · 10-16 weeks
Builder Path
Move from simulation and datasets to policies that can be tested on real or virtual robots.
Direct answer
Who is the Builder Path for?
Builder Path is designed for engineers and project teams and is structured as a 10-16 weeks sequence.
Course sequence
Complete the stages in order. Each page includes definitions, prerequisites, paper explainers, and recommended next steps.
- 01Open →
Foundations
The mathematics, programming, machine learning, vision, and robotics concepts needed to study Physical AI.
- 02Open →
Simulation
Virtual environments used to train, test, and evaluate embodied agents before real-world deployment.
- 03Open →
Datasets
Collections of robot trajectories, observations, actions, language instructions, and environment interactions.
- 04Open →
Data Annotation Pipelines
Human-in-the-loop and model-assisted systems that convert raw video, sensor streams, and robot logs into training data.
- 05Open →
Robot Data Flywheels
The loop of collecting demonstrations, mining failures, annotating trajectories, generating synthetic data, retraining policies, and redeploying robots.
- 06Open →
Robot Learning
Methods that allow robots to acquire behavior from demonstrations, rewards, interaction, or generated experience.
- 07Open →
Manipulation
Perception and control for grasping, moving, assembling, and using objects.
- 08Open →
Sim-to-Real Transfer
Techniques for transferring policies trained in simulation to physical robots.
- 09Open →
Deployment
Engineering reliable Physical AI systems under latency, compute, power, hardware, and operational constraints.
- 10Open →
NVIDIA Halos for Robotics
NVIDIA Halos is a full-stack functional-safety system for humanoids, industrial robots, and autonomous mobile robots; Halos OS is its operating-system layer, not a general robot-development OS.