Cloudera AI on Premises

Cloudera AI on premises delivers the agility and computational power of cloud computing directly inside your secure data center. Operating on OpenShift Container Platform or Cloudera Embedded Container Service, the platform delivers high-performance analytics with strict multi-tenant resource isolation and enterprise data governance.

To manage the complete lifecycle from data preparation to production inference, Cloudera AI is built on three core pillars:

  • Cloudera AI Workbench: provides isolated containerized environments for collaborative development and workflow automation.
  • Cloudera AI Registry: centralizes model versioning, MLflow artifact tracking, and fine-grained access control.
  • Cloudera AI Inference service: hosts real-time, auto-scaled model endpoints using vLLM engines and Multi-Instance GPU (MIG) resource virtualization to serve predictive models and Large Language Models (LLMs) with minimal latency.

Core capabilities and use cases

Data scientists, MLOps engineers, and developers can use Cloudera AI on premises across the following main enterprise use cases:

Interactive data exploration and analytics: Connect directly to enterprise data sources, including Cloudera Data Warehouse, Apache Spark, Apache Iceberg, Apache Ozone, Amazon S3, and SQL databases through JDBC, and launch containerized sessions using Python, R, or Scala ML Runtimes in JupyterLab or local IDEs such as VS Code.

Model training, distributed computing, and experiment tracking: Accelerate workloads using NVIDIA GPUs and RAPIDS runtimes, scale compute across clusters with the Workers API or YARN Spark pushdown, and record parameters, artifacts, and evaluation metrics using MLflow.

Workflow automation with jobs and pipelines: Schedule repeatable data transformation scripts, batch inference, or periodic retraining workflows using automated Jobs (with retry parameters) and multi-stage Pipelines.

Web application and dashboard deployment and sharing: Deploy interactive analytical applications built with frameworks such as Streamlit, Dash, Shiny, or Flask to share tools and visualizations directly with business stakeholders.

Enterprise model governance and lifecycle management: Register model artifacts using the UI or MLflow SDK, track training data lineage in Apache Atlas, enforce role-based authorization, and manage version transitions across environments.

High-performance REST API model serving: Host low-latency predictive models and LLMs supporting the OpenAI API protocol using vLLM engines, attach Low-Rank Adaptation (LoRA) adapters, and leverage MIG hardware slicing for multi-tenant isolation.