Praxis

Multimodal data infrastructure for Physical AI

Capture, curate, protect, govern, structure, and export.

Pipeline · end to end

Capture

RGB, depth, metadata

Curate

Time-aligned streams

Protect

PII redacted

Govern

Provenance and audit

Structure

Per-hand action data

Export

Training-ready episodes

Capabilities

Capture

Operator-side multimodal capture: RGB, depth, and session metadata.

Curate

Time-aligned streams. Slips and retries kept, not filtered.

Protect

PII redacted. Consent captured per episode.

Govern

Provenance and audit trail from raw frame to exported episode.

Structure

Per-hand action sequences with grasp, contact, and timing.

Export

Training-ready episodes for imitation and policy learning.

What Praxis enables

Scale data collection beyond the lab

Portable, operator-side rigs collect demonstrations in real environments, not just teleop on instrumented hardware.

Skip building the infra pipelines

Hands, depth, IMU, haptic, and more — sensor pipelines come built in.

Ship into regulated environments

Consent capture, PII redaction, and audit trail come with the data.

Spot data gaps before they cost a training run

Diversity, regional coverage, task balance — the metrics that decide policy quality, not just episode counts.

Structured action data · in depth

Structured action data from real-world manipulation.

InputCurateOutput

Observe

track hands and objects

Decompose

break into phases

Structure

structure action data

{
"task": "Pick the battery and place it down.",
"duration_sec": 3.65,
"segments": [
{
"action": "reach",
"language": "Right hand reaches for the battery.",
"object": "battery",
"right": {
"grasp_type": "precision",
"contact_state": "no_contact"
},
"t_start": 0.0, "t_end": 0.3
},
{
"action": "contact",
"language": "Right hand touches the battery.",
"object": "battery",
"right": {
"grasp_type": "precision",
"contact_state": "making_contact"
},
"t_start": 0.3, "t_end": 0.4
},
{
"action": "grasp",
"language": "Right hand picks up the battery.",
"object": "battery",
"right": {
"grasp_type": "pinch",
"contact_state": "making_contact"
},
"t_start": 0.4, "t_end": 0.6
},
{
"action": "translate",
"language": "Right hand slowly carries the battery.",
"object": "battery",
"right": {
"contact_state": "in_contact"
},
"t_start": 0.6, "t_end": 2.6
},
{
"action": "release",
"language": "Right hand slowly places the battery.",
"object": "battery",
"right": {
"contact_state": "breaking_contact"
},
"t_start": 2.6, "t_end": 3.5
}
]
}

Time-aligned action sequences. Per-hand grasp, contact, and timing.

Same pipeline across all episodes—consistent, training-ready.

See a slipInteractive · ~8s clip

From operator hands to training data.

For teams shipping robots into the real world. We’re working with a small number of design partners deploying Praxis in production.

Collaborate with us