DSCI-276: Cloudera AI Inference: From Model Deployment to Intelligent Applications
Overview
This 1-day instructor-led course provides a hands-on introduction to Cloudera AI Inference for deploying and managing LLMs models using AI Registry and scalable endpoints.
Participants will learn how AI Inference powers real-world use cases such as Cloudera Copilot, SQL AI Assistant, and multi-agent workflows with Agent Studio.
In addition, learners will explore key concepts such as model size and quantization, latency vs. throughput trade-offs, GPU selection, and workload sizing, enabling them to design cost-efficient and high-performance AI systems.
The training focuses on deploying and integrating scalable, secure, enterprise-ready AI capabilities using Cloudera.
What Skills You Will Gain?
Participants will develop the following skills:
Understand the architecture and core components of Cloudera AI Inference and AI Registry
Register, manage, and deploy ML models and LLMs as scalable inference endpoints
Configure CPU/GPU resources, and optimize performance using quantization and workload sizing
Understand key concepts such as latency, throughput, concurrency, and I/O ratios
Invoke and integrate inference endpoints using APIs and enterprise workflows
Enable AI-assisted development using Cloudera Copilot in JupyterLab
Use SQL AI Assistant to generate, optimize, and debug queries using natural language
Build and integrate AI-powered agents using Agent Studio for end-to-end workflows
Who Should Take This Course?
This course is ideal for AI/ML Engineers, Data Scientists, MLOps Engineers, Platform Engineers, Data Engineers, DevOps Engineers supporting AI workloads, Solution Architects, and Technical Innovation Teams responsible for deploying and operating AI models in enterprise environments.
Book the course
Course Details
Product Overview & Architecture
Overview of Cloudera AI Inference
Control plane and runtime architecture
Integration within the Cloudera AI ecosystem
CPU and GPU deployment models
Core enterprise capabilities
Overview of AI Registry
Cloudera Model Hub
Cloudera AI Registry — Register & Govern Models
Theory: Model Size & Precision
Model Size & Precision / Quantization
Latency vs. Throughput — The Core Trade-off
Theory: I/O Ratio & Concurrency
I/O Ratio & Concurrency — Sizing Your Workload
NVIDIA GPU Reference Guide for AI Inference
Theory: GPU Selection for AI Inference
GPU Selection for Cost Optimisation
NVIDIA NIMs & Benchmarking — Complexity to Confidence
Hands-On Exercise
Explore the AI Inference interface
Import an LLM into AI Registry
Deploy a Model Endpoint via the CAII UI
Authenticate & Call the Endpoint
Import Model via Hugging Face
Cloudera Copilot & AI-Assisted Development
- Overview of Cloudera Copilot
- Integration with JupyterLab editor
- Foundation on JupyterAI extension
- Connecting Copilot to models deployed in Cloudera AI Inference
Hands-On Exercise
Enable Copilot within AI Workbench
Connect Copilot to an AI Inference endpoint
Execute AI-assisted code generation
Loading the extension, listing models, registering aliases
Code generation with --format code for Python, PySpark
Output formats: markdown, html, math
Debugging SQL and Python errors with %ai error
Explaining complex code in plain language
Multi-turn iterative development with conversation context
Chat panel commands: /ask, /fix, /generate
Real-world challenges: ETL, ML templates, documentation
- Alias management: register, use, update, delete
SQL AI Assistant in Cloudera Data warehouse
- Introduction to SQL AI Assistance in Cloudera Data Warehouse and its key capabilities
- Natural language to SQL generation with support for Hive
- Foundation on JupyterAI extension
- Connecting Copilot to models deployed in Cloudera AI Inference
Hands-On Exercise
Enable AI Assist in Hue
Use GENERATE to create SQL queries from business questions
AI-powered query optimization and performance improvement techniques
Use FIX to debug and correct SQL errors
Use EXPLAIN to understand complex SQL queries in plain language
Multi-turn conversational interface using Assistant actions (Generate, Edit, Optimize, Fix, Comment)
Integrating Cloudera AI Inference with Agent Studio for Multi-Agent Workflows
Overview of AI Agents in enterprise automation and decision-making
Cloudera Agent Studio for building and managing multi-agent workflows
Integration with Cloudera AI Inference for real-time agent intelligence
Architecture of agent systems using LLMs, tools, and orchestration
Hands-On Exercise
Build and explore a Database Agent in Agent Studio
Connect the agent to Cloudera Data Warehouse (CDW) using Hive tools
Understand how the SQL Execution Agent interacts with database tools
Ask business questions in plain English using the agent interface
Automatically generate and execute Hive SQL queries on CDW
Observe multi-agent collaboration (SQL + Analysis Agent) for enhanced insights
Execute an end-to-end workflow from user query → SQL generation → data retrieval → response
Build confidence in using conversational AI for real-world data exploration