Keep the model and product yours
Dataplicity is not a training or annotation platform. The intended role is operating the model, runtime and resulting product workflow after deployment.
Operate AI workloads across the edge fleet.
Edge AI operations are undergoing validation. Operate an inference product across Linux devices: manage model versions and delivery, understand service health and connect inference results to the product workflow around them.
- Manage the deployed model as part of the product lifecycle.
- Keep model accuracy and application fitness with the team that owns the product.
- Use the same fleet operations around AI and non-AI workloads.
Operate the AI-powered product, not just the computer running it.
Cost-effective edge compute makes local inference increasingly practical. The operational question is how to deliver, observe and recover that workload across a fleet.
Fit it to your workload
Use the model, runtime and accelerator combination appropriate to your product, and keep the detailed integration choices in the technical guide.
Read the technical guideProve one workload on the hardware you intend to ship.
Start with the model, runtime and device combination that represents the product you are building.
Deliver and run the model
Verify the supported model reaches the device, loads in the intended runtime and produces the result your application consumes.
Observe degradation and recovery
Test restart, lost connectivity and the fallback path you intend to depend on. Keep degraded and unavailable states explicit.
Check fleet evidence
Confirm operators can see the active version and service health needed to distinguish a healthy inference product from an online Linux host.
Read the technical guideUse fleet operations around the workload you already have.
Edge AI is an additional operating path, not a requirement for using Dataplicity around an existing inference application.
Reach and diagnose the device
Use remote access, Pulse and available diagnostic evidence around the workload you run today.
Explore this pathRead the integration detail
Runtime, accelerator, model-format and release-specific requirements belong in the technical guidance.
Read the technical guideTake one workload from model to operated product.
Use the technical guide for runtime and hardware integration, then connect the workload to the fleet and product workflow around it.