$2,100 - $3,400
Buy the site package
Use this when your team needs exact-site data for training, fine-tuning, or integration in your own stack.
- Walkthrough media and camera poses
- Geometry and depth when available
- Rights, freshness, and package notes
For Robot Teams
Train your robot on the exact customer site before you visit.
Blueprint turns real customer facilities into site-specific world models your team can evaluate policies on, generate training data from, and fine-tune against — before anyone books travel.
Train and evaluate policies on the exact customer facility.
Generate site-specific data for fine-tuning before anyone books travel.
Compare releases on the same site so regressions show up early.
Start with proof
See the real site first, then decide how deep to go.
The public demo listing answers the first trust question: is this a real site? From there, decide whether to buy the site package for your own training pipeline or request hosted evaluation.
- A real facility captured once and used for training, evaluation, and data export.
- A package path for teams that need exact-site data in their own training pipeline.
- A hosted path for managed evaluations, release comparison, and dataset generation.
Public walkthrough from the live demo listing.
Two access paths
Buy the site package or run it with Blueprint.
Use the site package when you need exact-site data in your own training pipeline. Use hosted evaluation when you want Blueprint to run evaluations, generate data, and export results.
$16 - $29 / session-hour
Run a hosted evaluation
Use this when you want Blueprint to run evaluations, generate training data, and export results from the exact site.
- Repeatable runs on one exact site
- Rollout video, metrics, and failure review
- Raw bundles and dataset exports for tuning and adaptation
$50,000+
Plan a custom program
Use this for custom capture, exclusive training data, or managed evaluation programs around one facility.
- Custom capture scope
- Private or exclusive access
- Managed support for high-stakes work
Why this works
Policies trained on exact sites outperform generic sim.
MIT's RialTo and similar sim-to-real studies found 50-200% gains when teams trained on reconstructed target environments instead of generic scenes. Blueprint gives your team that exact environment.
Policy evaluation
Run your policy on the exact site layout, task lane, and handoff points. Know what breaks before the field week starts.
Site-specific fine-tuning
Generate rollouts, observations, and failure cases from the real facility and feed them back into training.
Release comparison
Run the same site after each autonomy update. Catch regressions before they reach a customer.
Data generation
Export site-grounded datasets, varied scenarios, and edge-case rollouts for offline training and adaptation.
Three steps from site to shipping.
Pick one real site
Start from the customer facility and workflow that actually matters for your next deployment.
Choose how to access
Buy the site package to train in your own stack, or request hosted evaluation for managed runs and exports.
Train and ship
Fine-tune policies, compare releases, and bring site-grounded evidence into your deployment decision.
No guessing about what is real, supported, or allowed.
The trust layer sits right next to the listing so technical buyers can decide without a long back-and-forth.
Real sites, not synthetic stand-ins
Every world model is built from real indoor capture of the actual facility your robot needs to work in.
Clear buying options
Every listing points to the same two choices: buy the site package or request a hosted evaluation.
Visible compatibility
Robot assumptions, export formats, privacy notes, and freshness details stay visible where you make decisions.