Skip to main content

NVIDIA-Certified Professional: Gen AI LLMs — Learning Library

This folder is a concept-first learning library for the NVIDIA-Certified Professional: Gen AI LLMs certification. We learn each topic deeply before shifting to exam-style preparation.

Chapter sequence​

  1. LLM Fundamentals — tokens, embeddings, transformer architecture, self-attention, model families, decoding.
  2. Prompt Engineering and RAG — prompting patterns, structured output, retrieval, hallucination reduction.
  3. Data Preparation — curation, dataset quality, tokenizer choice and vocabulary design.
  4. Model Optimization — quantization, pruning, distillation, KV cache and efficient inference.
  5. Fine-Tuning and Alignment — SFT, LoRA/PEFT, preference optimization, evaluation of adaptation.
  6. Evaluation — metrics, benchmarks, human evaluation, LLM-as-a-judge and error analysis.
  7. GPU Acceleration and Distributed Training — precision, memory, profiling, DDP/FSDP and parallelism.
  8. Deployment — containers, Kubernetes, batching, NVIDIA NIM and Dynamo-Triton concepts.
  9. Production Reliability — observability, versioning, retraining and incident analysis.
  10. Safety, Ethics and Compliance — bias, guardrails, red teaming and responsible AI.

How to use each chapter​

  1. Read the core lesson once without trying to memorize every term.
  2. Work through the mini-exercises in your own words.
  3. Use one conceptual video and one hands-on resource.
  4. Keep a short notes file with: what it is, why it matters, how to recognize it in a real system, and tradeoffs.
  5. Return later for a quiz and scenario-based practice.

Status​

  • Chapter 1: LLM Fundamentals
  • Chapters 2–10

The study guide supplied by Himakar is the topic map; these files expand it into a learning-first path.