A Smarter Way to Unlock Unstructured Data: SAS RAM in Action
16m + Hands-On Practice
Available in:
A Smarter Way to Unlock Unstructured Data: SAS RAM in Action
AIRAM : AIRAM3
The SAS Retrieval Agent Manager workshop teaches you how to build and automate retrieval-augmented generation (RAG) workflows (pipelines), covering essential steps like data ingestion from various unstructured sources, organizing data into collections, configuring embedding models and extraction tools, evaluating pipeline performance, and selecting optimal configurations.
You will learn to interact with data through a chat interface, create agents for advanced processing and automation, integrate external tools and systems through the Model Context Protocol (MCP), and set up workflow automations.
Overall, you will gain hands-on skills in building, evaluating, and automating enterprise retrieval-augmented generation solutions
Learn How To
- Learn to build and automate end-to-end RAG pipelines, including data ingestion, organization, vectorization, evaluation, and automation.
- Practice loading unstructured data (PDF, JSON, images, text, Word), organizing it into collections, and configuring embedding models and extraction tools.
- Evaluate pipeline configurations using manual and automated tests, interpret results, and select the best-performing setup.
- Interact with collections through a chat interface and create agents to automate querying and post-processing.
- Create agents for advanced data processing, automation, and integration.
- Set up workflow automation so that updates to sources trigger downstream processing and agent reinitialization, reducing manual effort.
Who Should Attend
- Technical Roles:
- AI/ML Engineers: Gain hands-on practice configuring embedding models, integrating LLMs, optimizing vector databases, and automating RAG pipeline operations — skills you can apply directly to production AI/ML workloads.
- Application Developers: Learn how to extend and integrate SAS RAM with other enterprise systems through APIs, agents, and automation tools, so you can ship RAG-powered applications faster.
- Business Roles:
- Business Analysts & Research Analysts: Learn to use the chat interface and built-in evaluation frameworks to discover and validate enterprise information yourself, without waiting on engineering support.
- Tangential Roles:
- Data Engineers: Understand how source ingestion, configuration, and vectorization choices affect downstream retrieval quality, so you can design and maintain data pipelines that feed reliable RAG systems.
- System Administrators: Learn to configure access, manage user permissions, and oversee deployment and security for SAS RAM, so you can support and govern the platform confidently within your organization.
Prerequisites
Before attending this course, you should be familiar with:
- Basic conceptual understanding of Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG).
- Comfort reading Python code (exercises use provided scripts; no need to write code from scratch).
- Basic command-line/terminal comfort (some exercises run scripts via bash and inspect output).
- Basic understanding of embedding models and vector search concepts.
SAS Products Covered
SAS Machine Learning;SAS Viya
Course Outline
This workshop provides hands-on experience in building and deploying RAG-based applications:
- Configuration
- Configure large language models (LLMs) and embedding models, and set up a vector database destination.
- Retrieve Source Data
- Use a custom script to query and download PDF files from arXiv, an open-access repository widely used for preprints in physics, computer science, and AI.
- Configure a Collection and Vectorize
- Define a configuration with an embedding model, chunking strategy, and optional OCR/table extraction.
- Vectorize the source documents and store the embeddings in a vector database for efficient retrieval.
- Evaluation Phase
- Create both user-driven and automated evaluations to validate the setup, and interpret Ragas metrics.
- Once the configuration meets quality standards, designate it as the Champion for production use.
- Chat
- Start interacting with the Retrieval-Augmented Generation (RAG) system for contextual responses.
- Agents
- Build a Code Template with the agent's logic, then create an agent and an Agent Experiment that ties them together and leverages the curated collection.
- Select the experiment as Champion before starting the agent, then chat with it to generate summaries tailored to specific requests.
- MCP Tools
- Create an MCP tool server template and an MCP tool server.
- Build an agent that leverages those tools alongside the curated collection.
- Automation Workflow
- Implement automation so that any change in the source triggers a vectorization job and reinitializes the agent.