A complete school of Vision-Language-Action (VLA): the technology that enables robots to see, understand, and act in the physical world. From fundamentals to deployment, in the year robotics had its "ChatGPT moment".
Market for US$ 38 bilhões até 2035 · +78 million new jobs by 2030 · 530,000 tech jobs without qualified professionals in Brazil
100% online · Start with the fundamentals · Immediate access to open content
US$ 38 bi
Humanoid market in 2035 Goldman Sachs, 2025
+101%
Jump in Physical AI investment in 2025 PitchBook
+78 million
Net new jobs by 2030 WEF Future of Jobs 2025
530,000
Tech jobs unfilled in Brazil 2025
The inflection point
In October 2024, Physical Intelligence launched the π0 — the first model trained at internet scale to make a robot fold laundry, assemble boxes, and make coffee without being programmed for each task. In January 2025, a BMW put VLA models into permanent production at a factory. In 2026, Jensen Huang (NVIDIA) stated: the turning point for robotics has already happened — isn’t coming.
The cost of a humanoid fell from US$ 85.000 para US$ 25.000 in two years. A research robotic arm, the SO-101 from HuggingFace, costs ~US$ 100. The best models — OpenVLA, π0, GR00T N1 — are open. For the first time in history, anyone can learn, train, and deploy AI that acts in the physical world.
The window is open. The question is: will you get in now — or in five years, competing with everyone else?
Market research · 2024–2026
Data from Goldman Sachs, Morgan Stanley, WEF, PwC, Crunchbase, PitchBook, and IFR. Every number here has a source and year.
Range of analyst projections. The gap is so wide it becomes an argument: no one knows the ceiling.
Morgan Stanley goes further: projects US$ 5 trilhões e 1 bilhão de humanoides em uso até 2050 (2024).
💰
VC investment in robotics jumped from US$ 7,8 bi (2024) para US$ 13,8 bi (2025) — +77% (Crunchbase). By the broader definition of "physical AI," US$ 27.6 billion, more than doubling in one year (PitchBook).
📊
The WEF projects 170 million new jobs and 92 mi displaced by 2030 — a net gain of +78 mi. 58% of employers say robotics and automation will transform their businesses (Future of Jobs 2025).
🧾
AI jobs pay a median of US$ 156.998 (PwC, 2025), and those with AI skills earn +56% that peers. Robot/Embodied AI Engineer in the US: US$ 122k a 335k/ano (Glassdoor, Meta, GM).
| Company | Valuation | When |
|---|---|---|
| Figure AI | US$ 39 bi | Sep 2025 |
| Skild AI | US$ 14 bi+ | Jan 2026 |
| Physical Intelligence (π) | US$ 5,6 bi | Nov 2025 |
| Apptronik | US$ 5,5 bi | Feb 2026 |
| NEURA Robotics 🇩🇪 | — | Mar 2026 |
Sources: TechCrunch, Bloomberg, Crunchbase News, AIFundingTracker (2025–2026). Jeff Bezos personally invested in 7 of these companies.
Global talent race
While governments invest billions in training, Brazil still has no specialized VLA training. This is the space this school fills.
530,000
tech professionals in short supply in 2025—for only ~53,000 graduates per year.
R$ 23 bi
from the Brazilian AI Plan (2024–28) for infrastructure and training.
0
specialized training in VLA / embodied AI in the domestic market.
SENAI-SP trains 1 million apprentices/year and has already launched an AI program with ServiceNow (Aug 2025) — but in general digital skills, not in intelligent robotics. Translation: the demand exists, the money exists, and no one is training this professional yet. Those who learn now will get ahead before the market matures.
The school
A curriculum in 3 learning paths and 18 modules — the same concepts taught at Stanford, Berkeley, and CMU, in Portuguese and with accessible open-source hands-on practice. Get started right now for the open content.
The evolution from LLM → VLM → VLA, model anatomy, an overview (RT-2, OpenVLA, π0), datasets, hardware, and safety (ISO 10218).
6 modules · open learning path →Python, PyTorch and HuggingFace, end-to-end LeRobot, simulators (MuJoCo, Isaac Sim), ROS2, training (ACT, Diffusion) and sim-to-real.
6 modules · open learning path →Dual-system architecture, world models, cross-embodiment generalization, industry applications, the US$ 7.2 billion market, and the roadmap through 2030.
6 modules · open learning path →Open and free content · 50+ academic and industry sources
Where the technology you'll learn is already creating value in the real world — with named companies.
🏭
BMW has run VLA models in continuous production since Jan/2025. Tesla Optimus works in its own factories, targeting 1 million units per year by 2027.
📦
Amazon Robotics uses the Digit humanoid (Agility) in distribution centers. Skild AI handles inspection, last-mile delivery, and warehouses.
🏥
A CommonSpirit Health invested in Skild AI (Jan/2026). Physical Intelligence tests household assistance and care tasks.
🌾
Selective harvesting, phenotyping, and vision- and language-guided spraying — one of the target sectors in Japan’s and China’s robotics policies.
🔓
LeRobot (HuggingFace) + arm SO-101 at ~US$ 100 make it possible to train a real robot at home. The NVIDIA Physical AI Dataset has already surpassed 4.8 million downloads.
👷
In China, the work of teach robots through teleoperation. Robot Learning Engineer, Sim Engineer, and Data Teleoperator are roles that didn’t exist 3 years ago.
You’re a developer or engineer and wants to move into the fastest-growing and highest-paying field in AI.
Study AI/ML but feels like everything is text and images—and wants to enter the physical world.
Works in robotics/automation and wants to master the intelligence layer (VLA).
You’re an entrepreneur or leader and needs to understand in depth where intelligent robotics will create business opportunities.
Honest prerequisite: programming logic and a willingness to learn. The school teaches the rest, from fundamentals to practice.
Start today with the open fundamentals and secure your spot in the first cohort of the full program. The market has a US$ 38 billion ceiling, and Brazil has no specialized training. You decide which side of the curve you’ll be on.
No fluff: the curriculum is already online. You can study now and decide later.