Home / Quality & formats

Data your training run can trust.

Raw capture is rarely training-ready. We clean it, score it, and hand it back in the exact schema your pipeline expects — with the audit trail to prove nothing was lost on the way.

CLEANING & QA

Find the failures before your model does.

We catch dropped frames, desynced sensors, label drift, duplicates, and outliers, then validate every episode against task-level rules — each ships with a quality score.

  • sync validation
  • outlier detection
  • label audit
  • dedup / balance
  • per-episode scoring
FORMAT CONVERSION

Move between standards without loss.

We map schemas explicitly, preserve modalities and timestamps, and verify every conversion with an automated round-trip — so datasets never silently degrade into your training loop.

  • RLDS
  • HDF5
  • MCAP
  • LeRobot
  • ROS bag
  • custom schemas
Speak your standard

Any direction, verified round-trip.

Datasets rarely arrive in the format you train on. We move between them — and check the trip both ways.

from \ toRLDSHDF5MCAPLeRobot
RLDS
HDF5
MCAP
LeRobot

Your schema, preserved

Every conversion maps fields explicitly, keeps timestamps and modalities intact, and is round-trip-checked so nothing silently degrades.

Working in something custom? We map it. ROS bags, Parquet, Zarr, and bespoke episode schemas are all in scope.

RLDSHDF5MCAPLeRobotROS bagParquet / Zarr
Delivery & provenance

Auditable from sensor to shipment.

SCORED

Quality metrics included

Every batch ships with pass rates, agreement stats, and per-episode quality scores — you see the numbers, not just the data.

VERSIONED

Full dataset lineage

Every cleaning rule, conversion, and label revision is versioned. You can trace any episode back to its raw capture.

VERIFIED

Round-trips checked

Conversions are validated by converting back and diffing — timestamps, modalities, and schema fields all confirmed intact.

RULES

Your task logic enforced

Task-level validation rules — gripper states, contact events, workspace bounds — are encoded and applied to every episode.

SECURE

Your data stays yours

Strict access controls and NDAs by default. Data is processed in isolated environments and deleted on your schedule.

FAST

Pilot in days

Send a sample dataset and get it back cleaned, scored, and converted within 48 hours — judge the quality yourself.

Try the pipeline

Send us a messy dataset.

Share a sample of raw episodes and your target format. We'll return them cleaned, scored, and converted — with the QA report — so you can judge the work yourself.

Request a pilot → or email contact@intellisant-ai.com