NVIDIA NCA-AMAD: NVIDIA-Certified Associate: AI Models and Application Development

Last checked October 22, 2026 against the official NVIDIA exam page.

1. Question 1Machine Learning Fundamentals

   A team has labeled data of house sizes and their sale prices and wants to predict the price of a new house. Which type of learning is this?

   1. AUnsupervised clustering
   2. BAnomaly detection with no labels
   3. CSupervised regression
   4. DReinforcement learning
   Show answer and explanationHide answer

   Correct answer: C

   Predicting a numeric value from labeled examples is supervised regression. Clustering works without labels, reinforcement learning uses rewards, and anomaly detection looks for unusual points.

2. Question 2Transformers and Language Models

   What is the main purpose of the attention mechanism in a transformer?

   1. ACompressing the entire input sequence into a single fixed-size number before the decoder layers process it
   2. BStoring the full vocabulary and token-to-ID mapping in an external database for lookup at inference time
   3. CLetting each token weigh the relevance of other tokens when building its representation
   4. DRemoving punctuation and special characters from the raw input text during preprocessing stage
   Show answer and explanationHide answer

   Correct answer: C

   Attention computes weighted relationships between tokens so each representation can use context. Compression to one number, punctuation removal and database storage are not its purpose.

3. Question 3Model Deployment and Inference

   A deployed model must be monitored for quality degradation over time. Which signal is most useful?

   1. AThe number of times the inference server was restarted in the past quarter
   2. BA periodic evaluation on a fixed test set and tracking of user feedback
   3. CThe color of the server rack's front door in the data center hall
   4. DThe length of the model's name including its version tag
   Show answer and explanationHide answer

   Correct answer: B

   Periodic evaluation on fixed tests and user feedback reveal quality changes. Restart counts, rack colors and name lengths do not measure quality.

4. Question 4Prompt Engineering

   A prompt includes several worked examples of the desired input and output before the real input. What is this technique called?

   1. AZero-shot prompting
   2. BFew-shot prompting
   3. CGradient clipping
   4. DModel distillation
   Show answer and explanationHide answer

   Correct answer: B

   Few-shot prompting shows examples in the prompt to guide the output format and behavior. Zero-shot provides no examples, and clipping and distillation are training techniques.

5. Question 5Retrieval-Augmented Generation

   In a retrieval-augmented generation system, what is the role of the vector database?

   1. ATraining the language model from scratch on the retrieved passages
   2. BStoring embeddings of document chunks and returning the most similar chunks for a query
   3. CGenerating the final answer text token by token with sampling after reading the retrieved chunks
   4. DStoring the user's password hash and session tokens with encryption keys
   Show answer and explanationHide answer

   Correct answer: B

   A vector database stores embeddings and retrieves similar chunks for the query. Training and final generation are performed by the model, and passwords are not stored there.