FAQ

Q1. Can I create multiple pipelines under the same project? Yes. You can create and manage multiple pipelines within the same project, each with its own configurations, datasets, and models.

Q2. Can I run multiple pipelines simultaneously? Yes, if sufficient system resources are available. However, heavy workloads may impact performance, so it’s recommended to monitor resource usage in the dashboard.

Q3. Can I select multiple models to be trained in one pipeline execution? Yes. You can configure the training step to include multiple classification and regression models, and the platform will train them in parallel depending on your resource allocation.

Q4. Can I compare multiple models trained in the same run? Yes. The experiment run results page allows side-by-side comparison of metrics, parameters, and performance for that run.

Q5. Can I version-control my transformed dataset? Yes. Dataset versioning is built-in, after every successful pipeline execution the data is versioned automatically.

Q6. Can I use external data sources like S3, GCS, or databases? Yes. The platform supports multiple external connectors for fetching data from Azure, S3, Databricks, Snowflake and Postgres.

Q7. Can I deploy models as REST APIs? Yes. The deployment module allows you to publish models as REST endpoints for real-time predictions.

Q8. Can I deploy multiple models from the same project? Yes, you can deploy multiple models to different endpoints.

Q9 Is it mandatory to select additional algorithms? No. Selecting extra algorithms is optional — you can proceed with the default algorithm settings.

Q10. What does it mean if a model’s status is “Registered”? It means the model has been saved to the Model Hub but is not yet deployed for consumption. You must deploy it before it can receive prediction requests.

Q11. What is Online Serving? Online Serving lets you send a single data point for immediate prediction. It’s ideal for quick validation and sanity checks.

Q12. When should I use Online Serving vs Batch Serving? Use Online Serving for quick, one-off predictions; use Batch Serving for large-scale, scheduled predictions on big datasets.

Q13. Can I reuse my original training pipeline in Batch Serving? Yes. You can select the original training pipeline and optionally include pipeline steps for data transformation before predictions.

Q14. Where are Batch Serving results stored? The platform stores prediction outputs in the specified data storage location for later review or integration into downstream applications.

Q15. How does the platform detect drift? It compares new cleaned data (from pipeline executions) against the original training data baseline. Significant changes are flagged in the relevant dashboard.

Q16. What actions can I take if drift is detected? You can retrain the model, adjust preprocessing steps, or investigate data anomalies in your sources.

Q17. Can I monitor both classification and regression models? Yes. The module provides dedicated dashboards for both classification and regression performance tracking.

Q18. What are access roles and functional roles?

• Access Roles define which credentials a user can use in the platform.
• Functional Roles define which actions (read, write, delete) a user can perform in specific modules.

Q19. Who can create and assign roles? Only the Org Admin can create, grant, and assign both access and functional roles.

Q20. Can a user have both access and functional roles? Yes. Users can be assigned both types of roles, which together determine their platform permissions and credential usage.