Enrolling for 2026 Cohort

World Model
Certifications

From young explorers to expert engineers — a complete credential pathway for building AI systems that understand, predict, and reason about the physical world. From the Castalia Institute.

7
Credentials
190+
Lab Hours
Ages 10+
Rover Pathways
Open
Enrollment
Scroll
Rover Pathways · Young Learners

Start the journey early

The Rover Pathways are designed for younger learners who want to explore how the world works and build toward a career in world model engineering. Each pathway feeds directly into the professional certification tracks.

Missions & simulators at rover.castalia.institute
Lab hardware powered by terrAIn · terrAIn Home kit
Explorer
WM-010
Ages 10–13

World Explorer

Observe, hypothesize, and experiment. Drive real RC vehicles on the terrAIn bed, measure forces with sensors, and discover how the physical world works through hands-on missions.

How Things Move: Forces & Motion
Rover Sensors: Seeing the World Like a Robot
terrAIn Lab: Drive & Observe on Real Soil
Build Your First Simulation in Scratch
16 hours Learn more →
Navigator
WM-020
Ages 13–16

Reality Navigator

Go deeper into 3D environments, game engines, and basic programming. Program an ESP32 rover on the terrAIn bed and build a digital twin of the terrain.

3D Worlds: Intro to Game Engines & Physics
Python for Young Engineers
terrAIn Lab: Program the ESP32 Rover
Digital Twins: Mirror the Real Terrain Bed
24 hours Learn more →
Cadet
WM-030
Ages 16–18

Simulation Cadet

Bridge to the professional track. Intro to machine learning, neural networks, and building simple world models. Train a rover to navigate the terrAIn bed autonomously.

Neural Networks from Scratch
Intro to Reinforcement Learning
Building a Simple World Model
terrAIn Capstone: Autonomous Excavation
32 hours Learn more →
Complete Learning Pathway

From curious explorer to world model engineer

WM-010 Explorer 10–13
WM-020 Navigator 13–16
WM-030 Cadet 16–18
Bridge to Professional Tracks
WM-100 Foundation
WM-200 Practitioner
WM-300 Specialist
WM-400 Expert
Professional Tracks · WM-100 – WM-400

Four levels of mastery

For graduates of the Rover Pathways or experienced practitioners. Each certification builds on the last, taking you from foundational theory to production-grade world model engineering.

Foundation
WM-100

World Model Foundations

Core principles of world models — from mental simulation and intuitive physics to learned latent dynamics. Build your first predictive environment model.

Intuitive Physics & Object Permanence
Latent State Representations
Generative Environment Models
Evaluation & Benchmarking
30 hours Learn more →
Practitioner
WM-200

Simulation Intelligence

Design and train neural simulators for robotics, autonomous systems, and scientific discovery. Master differentiable physics and sim-to-real transfer.

Differentiable Physics Engines
Neural Radiance & Gaussian Splatting
Sim-to-Real Transfer on terrAIn
Multi-Modal Sensor Simulation
40 hours Learn more →
Specialist
WM-300

Embodied AI Architect

Build agents that learn and act within world models. Covers model-based reinforcement learning, planning through imagination, and real-time embodied systems.

Model-Based RL & Dreamer Architectures
Planning Through Imagination
terrAIn Multi-Vehicle Coordination
Real-Time Embodied Systems
35 hours Learn more →
Expert
WM-400

World Model Engineer

The capstone credential. Design production-scale world models for industry applications — from autonomous driving to surgical robotics to climate simulation.

Production-Scale Architecture
Distributed Training & Inference
Safety & Alignment for World Models
terrAIn Capstone: Autonomous Earthmoving
45 hours Learn more →
terrAIn · Hands-On Lab Platform

Where the curriculum meets real dirt

Every credential is backed by lab work on terrAIn — an open-source robotics platform where RC construction vehicles learn to dig, grade, and haul on real terrain. A 4x8 ft bed of soil becomes a world model laboratory.

terrain.castalia.institute

terrAIn labs across the pathway

WM-010 Explorer

Drive a real rover, observe how soil behaves, measure forces with sensors

WM-020 Navigator

Program an ESP32 rover in Python, build a terrain map from camera data

WM-030 Cadet

Train a neural network to navigate the terrain bed autonomously

WM-200 Simulation

Build an Isaac Sim digital twin of the terrain bed, transfer policies to real hardware

WM-300 Embodied AI

Multi-vehicle coordination — excavator loads truck while bulldozer grades

WM-400 Capstone

Design and deploy a production-grade autonomous earthmoving system

terrAIn Home

$599 – $899
Rover Pathways (WM-010 – WM-030)

One bed, one vehicle, base station, and sensors. Everything a family or co-op needs for the youth credentials. Real dirt, real robots, real learning.

Single RC vehicle (excavator or truck)
ESP32-S3 + micro-ROS firmware
4×8 ft terrain bed plans
Overhead camera setup
All Rover Pathway lab exercises

terrAIn Research

$2,000 – $4,000
Professional Tracks (WM-100 – WM-400)

Full fleet, Jetson base station, perception gantry, ROS2 stack, and Isaac Sim integration. The lab platform for serious world model engineering.

Full fleet: excavator, truck, dozer, loader
Jetson Orin Nano + ROS2 stack
Perception gantry with RGB + depth
Isaac Sim digital twin scenes
Multi-agent coordination labs
Capstone project infrastructure

The terrAIn stack

The Bed
4×8 ft soil, sand, or gravel
🚜
The Fleet
Huina 1:14 RC vehicles + ESP32
📷
The Eye
Overhead RGB + depth cameras
🧠
The Brain
Jetson Orin Nano + ROS2
🖥️
The Sim
NVIDIA Isaac Sim digital twin
Curriculum

26 weeks to mastery

A rigorous, lab-intensive program covering the full stack of world model engineering.

Phase 1 Weeks 1–6

Foundations

What is a World Model? — From Cognitive Science to AI
Representation Learning for Physical Environments
Autoregressive vs. Diffusion-Based Generation
Intuitive Physics: Gravity, Collisions, Fluids
Benchmarking World Models: Metrics That Matter
Lab: Train a Video Prediction World Model
Phase 2 Weeks 7–14

Simulation & Perception

Neural Scene Representations (NeRF, 3DGS)
Differentiable Rendering & Physics
Multi-Sensor Fusion: Vision, LiDAR, Proprioception
Generative Simulation for Synthetic Data
Sim-to-Real Gap: Domain Randomization & Adaptation
Lab: Build a Differentiable Simulator (terrAIn digital twin)
Phase 3 Weeks 15–20

Agents & Planning

Model-Based Reinforcement Learning (MBRL)
Dreamer, IRIS, & Imagination-Based Planning
Hierarchical World Models for Long-Horizon Tasks
Language-Grounded World Models
Safety-Constrained Planning in Learned Simulators
Lab: Train an Agent That Dreams Before It Acts (terrAIn fleet)
Phase 4 Weeks 21–26

Production & Capstone

Scaling World Models: Architecture & Infrastructure
Distributed Training on GPU Clusters
Real-Time Inference & Edge Deployment
Alignment & Safety for Predictive Models
Industry Applications: Robotics, AV, Climate, Medicine
Capstone: End-to-End World Model System (terrAIn deployment)
Why World Models

The models that understand reality

Large language models predict tokens. World models predict physics. They are the foundation for AI that can truly understand and interact with the real world.

The Next Frontier of AI

World models are how AI systems will understand causality, predict outcomes, and reason about the physical world. Every major AI lab is investing here.

Industry Demand

Autonomous vehicles, robotics, drug discovery, and climate science all require engineers who can build and deploy world models at scale.

Rigorous & Hands-On

Every module includes lab work. You won't just read papers — you'll implement world models from scratch, train them on GPU clusters, and deploy them.

Recognized Credential

Castalia Institute certifications are recognized across the AI industry. Join a global network of world model engineers building the future.

Ready to build the future?

The 2026 cohort is now open. Join engineers from leading AI labs, robotics companies, and research institutions.