Iristica Labs
Robotics training data
Robotics data platform
Robotics models fail without clean labels. Iristica Labs delivers consistent, human-verified annotations for perception and manipulation so your systems work in the real world.
Data quality
Peer-reviewed
Throughput
High-volume labeling
Delivery
Versioned datasets
Active labeling queue
LiveBin pick - success state
QC in progress
0.22 SOL
Drawer close - phase tags
Ready to label
18 USDC
Grasp retry - failure
Queued
0.12 SOL
Payout routing example
Split payouts across wallets
How it works
Each task is labeled and reviewed before it ships. Approved work is bundled into versioned releases so buyers can trust the data.
Step 01
Ingest
Rights-cleared robotics footage enters the queue.
Step 02
Label
Labelers create structured, consistent annotations.
Step 03
QC
A second pass review catches errors and drift.
Step 04
Payout
Approved work is paid quickly and transparently.
Step 05
License
Clean data ships as versioned releases.
Quality at scale
We combine human review, clear taxonomies, and performance tracking to keep quality high as volume grows.
Human-in-the-loop
Every task is labeled and verified by real people.
Consistent taxonomy
Labels follow a clear, enforced schema across releases.
Quality tracking
Review outcomes drive performance scores and access.
Performance boosts
High-quality work earns priority and bonuses.
Payout routing
Payouts can be split across worker pools and ops.
Auditability
Every release is traceable to its source tasks.
Datasets
Robotics teams get consistent labels, clear taxonomy, and version history. We also build custom collections on spec.
Rights cleared
Only licensed or first-party footage enters the pipeline.
Quality control
Every batch is reviewed before it ships.
Versioned delivery
Every dataset is versioned and reproducible.
Business model
The core business is dataset licensing and managed collections, with a transparent payout layer for labelers.
Dataset licensing
Clean, versioned datasets are licensed to robotics teams.
Managed collections
Custom, on-spec datasets for new robot platforms.
Flywheel
Better labels create better robots. Better robots create more demand for data.
Roadmap
We begin with a focused task domain, prove quality and throughput, then expand to new robots and buyers.
Phase 1: Seed pipeline
Ingest videos, recruit labelers, ship v1.
Phase 2: Fee routing
Route creator fees to payouts and QA.
Phase 3: Licensing scale
Expand task coverage and ship more releases.
For buyers
Robotics teams can license existing releases or request new task coverage with guaranteed quality.