Roma Tech AI is the education and tech-media studio for the era of embodied intelligence. Deep courses, original reporting, and weekly dispatches on physical AI and autonomous driving.
Trusted by engineers from Waymo, Tesla, NVIDIA, Figure, Wayve & Cruise
Roma Tech AI focuses on the two technology stacks reshaping the physical world. Everything we publish — courses, dispatches, interviews — sits inside these tracks.
From sensor fusion to whole-body manipulation — we unpack how today's humanoids, robot arms, and field robots actually think and act.
The full AV stack — sensors, planners, prediction, end-to-end driving — broken down by engineers who have shipped to production fleets.
Every course is co-authored by working engineers — from Waymo, Wayve, Figure, NVIDIA, and Tesla — and updated as the field moves.
A conceptual map of perception, planning, and control across robots and self-driving cars.
From multi-camera calibration to occupancy networks — implement a modern BEV pipeline in PyTorch.
Train and evaluate a tokenized driving policy — covering data curation, RL fine-tuning, and closed-loop sim.
MDPs, behavior cloning, and diffusion policies — the math and the code, side by side.
MPC, RL, and hybrid stacks. Hands-on with MuJoCo and Isaac Lab; deploy a walking policy.
Pretraining, post-training, and Sim2Real for VLAs — built around π0, OpenVLA, and friends.
Weekly stories on the companies, papers, and engineers building physical AI and self-driving.
A frame-by-frame teardown of the new end-to-end stack — and what its driving behavior reveals about its training data.
An hour with Figure's perception lead on data, reward shaping, and the company's bet on a single foundation model.
Occupancy is eating BEV. Here's what changed in 2026, who's leading, and how to evaluate one.