Understand RAG architecture - You will learn how to design data flow in a RAG system and clearly distinguish it from a standard LLM setup, so you can choose the right components for your own use cases.
Prepare data for retrieval - You will learn how to extract content from documents, split it into meaningful chunks, and tune chunking parameters so your downstream retrieval works more accurately.
Choose the right embeddings - You will compare embedding models and learn how to select vector representations based on data type, language, and business goal instead of relying on trial and error.
Use Qdrant confidently - You will configure the environment, create collections, and practice record operations in Qdrant, giving you a solid backend for applications built on vector retrieval.
Improve retrieval relevance - You will combine semantic search, metadata filtering, and reranking to return more relevant documents and reduce the number of weak or misleading answers from the system.
Build stronger prompts - You will practice creating prompt templates and combining user queries with retrieved context, so the model produces source-grounded answers instead of guesses or hallucinations.
Evaluate RAG quality - You will work with metrics and tools, including deepeval, that help you test answer quality, measure system performance, and automate validation as your RAG solution evolves.
Leave with a working blueprint - You will see a complete application for feeding a knowledge base and generating answers, making it much easier to transfer the techniques into a pilot or production project.
RAG architecture is presented as a way to extend language models with external knowledge sources. The section explains how the main components interact, how information moves through the system, and where LangChain supports the construction of end-to-end applications.
Before retrieval can work well, documents must be converted into usable input units. The focus is on extracting content with parsers, dividing text into chunks, and selecting splitting parameters that preserve meaning while supporting efficient indexing and later search.
Embeddings convert textual meaning into vector form, which enables semantic comparison and retrieval. This section explains how current embedding models and generators differ, and how to choose a representation approach that matches the nature and variability of the processed data.
Qdrant serves as the storage layer for vector data and related records in the RAG pipeline. The training addresses environment setup, collection management, record operations such as add, update, and delete, and practical integration of Qdrant with LangChain components.
Finding the right document depends on more than a basic similarity search. Participants work with semantic retrieval, metadata-based filtering, and reranking techniques to improve result precision and increase the relevance of materials passed to the generation stage.
Prompt Templates shape how the model receives instructions and uses retrieved evidence. The section focuses on combining user queries with contextual materials and on generating responses that synthesize source content instead of relying on unsupported model assumptions.
Reliable RAG output requires systematic evaluation, not only manual review. This part uses the deepeval library to define and measure key quality metrics, and to automate verification of whether generated answers remain correct against the expected standard.
A complete application is used here to connect the earlier stages into one working RAG flow. Participants follow how data is loaded into the database and then turned into usable knowledge through response generation based on the prepared resources.
Python basics - You should be comfortable reading and writing simple Python code, using functions, imports, and text processing, because the training focuses on building working solution components.
API and library usage - You should have experience installing libraries, working in a development environment, and using APIs, so you can run the examples smoothly and integrate the components.
LLM fundamentals - You should understand what language models, prompts, and response context are, so you can move quickly into RAG, embeddings, and grounded answer generation.
Data and search basics - You should know the basics of working with text data, documents, and simple information retrieval, since the training covers parsing, chunking, and retrieval workflows.
