Model Serving Platform
Model Serving Platform lets you manage, deploy, and serve proprietary models on your GPU clusters.
Before Activation
- Register and operate GPU clusters and resource groups.
- Manage models, deployments, inference, datasets, training, and evaluations.
- Observe workload health, capacity, and serving behavior.
- Integrate through the SDK and CLI and follow product troubleshooting guidance.
Explore the Documentation
Model Serving Workflow
Model Serving Platform supports the following model serving lifecycle:
- Register and manage proprietary model assets.
- Prepare datasets and configure training or evaluation jobs when needed.
- Deploy models as managed model services.
- Manage model versions and serving configurations.
- Monitor cluster resources and model service health.
- Integrate model services into internal or customer-facing platforms.
Core Capabilities
The following table summarizes the main platform capabilities:
| Capability | Description |
|---|---|
| Proprietary Model Management | Manage proprietary model assets in a centralized workspace. |
| Inference-Optimized Models | Deploy and host models optimized for inference workloads. |
| Model Deployment | Deploy proprietary models as managed services with version management. |
| Model Training | Manage datasets and run model training workflows. |
| Model Evaluation | Evaluate model quality before serving a model version. |
| Cluster Monitoring | Monitor GPU and compute cluster resource usage. |
| Model Service Monitoring | Monitor the health and status of running model services. |
Model Management and Serving
Model Serving Platform provides a centralized workflow for managing proprietary models and turning them into usable model services.
- Manage model assets and versions.
- Deploy models to available GPU or compute resources.
- Configure and operate model services.
- Update or manage deployed model versions.
- Monitor the status and health of model services.
See Deploy Models for deployment workflows.
Training and Evaluation
The platform supports the model development workflow before deployment.
- Prepare and manage datasets.
- Run model training jobs.
- Evaluate model quality.
- Use evaluation results to decide which model version to serve.
See Train and Evaluate Models for the complete workflow.
Platform Integration
Model services can be used for internal AI workloads or integrated into existing platforms for monetization.
The platform can be operated through the product console, SDK, or CLI.
Runtime operations require an activated service and a deployed Model Serving Platform instance. Documentation remains publicly readable before activation.