Skip to main content
Now let’s get started on the course for the NVIDIA Generative AI LLMs Associate Certification. We will begin with Core Machine Learning and AI Knowledge. Consider this question: A developer is tasked with implementing a retrieval-augmented generation, or RAG, system. Which of the following is NOT a typical component of a RAG architecture?
  • A vector database for storing document embeddings
  • A text generation model for producing responses
  • A gradient descent optimizer for continual learning
  • A document chunking mechanism for preprocessing
Answer: A gradient descent optimizer for continual learning. Why this is the correct choice Retrieval-augmented generation (RAG) integrates an external retrieval step with a generative language model. The usual RAG pipeline focuses on fetching relevant context and conditioning a generator on that context; it does not perform live gradient-based training during inference. Typical components include:
  • Retrieval — finding relevant documents, passages, or knowledge snippets using embeddings and similarity search (commonly backed by a vector database).
  • Preprocessing / Chunking — splitting long documents into model-friendly passages or chunks so the retriever and generator can work effectively.
  • Generation — a language model (decoder, encoder-decoder, or instruction-following LLM) that conditions on retrieved context to produce the final response.
Quick overview table
The image depicts a flowchart with three stages: Retrieval, Chunking, and Generation, each connected by arrows.
Note on continual learning and gradient descent
  • RAG systems typically keep the generation model fixed at query time and rely on an external retrieval index to surface current knowledge.
  • Continuous model updates via gradient descent are not part of the inference pipeline. When you need the model to learn from new data, typical approaches include periodic offline fine-tuning, parameter-efficient updates (e.g., adapters), or re-embedding the index and refreshing the vector database.
  • In short: keep retrieval fast and up-to-date; apply training and weight updates as separate, controlled offline processes.
RAG architectures center on retrieval, preprocessing/chunking, and generation. Continuous weight updates via gradient descent during inference are not part of the standard RAG pipeline—updates are handled separately when needed.

Watch Video