🇧🇷 Building in Brazil for physical AI

Physical-world data
for embodied AI
in Latin America.

Motion Lab builds mobile data collection infrastructure to generate localized POV datasets for robotics, AI, and multimodal model companies.

Talk to Motion Lab arrow_forward View Trampo Beta
Three.js Earth data network
Drag to rotate
Sao Paulo nodeBrazilian physical-world data acquisition.
Global deliveryStructured datasets for AI and robotics teams.
Consent-firstCollector screening, task approval, and validation.
Data Layer Brazilian Context

Physical-world datasets we are building

Localized egocentric data from everyday Brazilian environments, designed for AI, robotics, and multimodal model development.

POV shot of human hands washing a white ceramic plate in a Brazilian kitchen
Kitchen_BR_01

Washing Dishes

Multi-object manipulation including glassware, cutlery, and local ceramic textures under varying suds conditions.

BETA DATA CATEGORY add_circle
POV close-up of hands folding a soft shirt in a Brazilian apartment
Laundry_BR_04

Folding Clothes

Deformable object manipulation focusing on edge alignment and textile interaction physics in domestic settings.

BETA DATA CATEGORY add_circle
Action view of someone sweeping dust on a Brazilian apartment floor with a broom
Surface_BR_02

Sweeping the Floor

Large-scale workspace interaction and reach-trajectory data across common Brazilian floor materials.

BETA DATA CATEGORY add_circle
POV hands arranging everyday household objects on a clean table
Objects_BR_07

Arranging Objects

Spatial reasoning data from sorting, grouping, aligning, and repositioning everyday household objects.

BETA DATA CATEGORY add_circle
First-person conversation scene at a home table for AI training data
Social_BR_03

Conversation Data

Multimodal interaction data for speech, gestures, turn-taking, gaze, and real domestic social context.

BETA DATA CATEGORY add_circle
south_america

Geographic Diversity

Built to capture regional variation across Brazilian households, objects, lighting, surfaces, routines, and physical environments.

high_quality

Neural Precision

POV video and structured metadata designed to support higher-quality training, evaluation, and validation workflows for physical AI.

verified_user

Ethically Sourced

Collector participation is designed around consent, quality validation, task approval, and fair compensation for accepted missions.

Our Systematic Pipeline

From task design to validated physical-world datasets for AI and robotics teams.

01

Task Definition

Defining useful real-world interactions based on robotics needs and Latin American environmental context.

02

BR Network Record

Brazilian contributors receive guided missions through Trampo and record POV interactions with everyday objects and spaces.

03

Neural Audit

Quality checks review instruction adherence, visual clarity, safety, and dataset usefulness before approval.

04

Global Delivery

Approved data can be structured, labeled, and delivered to AI, robotics, and multimodal model companies.

Brasil • Inovação • IA

The Future of Robotics is
Trained in Brazil.

Talk to Gabriel View Trampo Beta