NVIDIA-Certified-Professional Accelerated Data Science Sample Questions:

1. You are analyzing a large financial dataset containing stock market tick-by-tick data stored in a cuDF DataFrame. Since the dataset contains billions of data points, you need to aggregate it at the minute level before visualizing price trends efficiently.
Which of the following is the best approach for aggregating and visualizing this time-series data using NVIDIA technologies?

A) Use cuDF's .groupby() function to aggregate at the minute level, then visualize using hvPlot
B) Load the data into a relational database (e.g., PostgreSQL), run an SQL query for aggregation, and visualize using Seaborn
C) Use cuML's TSNE function to reduce dimensionality before visualizing with Bokeh
D) Convert cuDF to Pandas, aggregate using .resample() in Pandas, and visualize using Matplotlib

2. A machine learning engineer wants to evaluate the performance of NVIDIA RAPIDS cuDF and Apache Spark for large-scale data processing on a GPU-enabled cluster.
Which of the following strategies is the most effective for obtaining a fair and comprehensive benchmark?

A) Run Spark on a CPU cluster while running RAPIDS on a GPU to compare real-world scenarios.
B) Focus only on processing speed without considering resource consumption differences between frameworks.
C) Execute identical ETL workflows on cuDF and Spark-RAPIDS and measure execution time and resource utilization.
D) Limit the benchmark to small datasets since GPUs excel at parallel processing.

3. A data scientist is working on a machine learning model for fraud detection. Due to the limited size of the dataset, they decide to generate synthetic data using NVIDIA RAPIDS AI and cuDF.
Which of the following approaches is the most efficient and effective for generating synthetic data while ensuring compatibility with RAPIDS AI workflows?

A) Train a generative adversarial network (GAN) using PyTorch and then use the generated samples in RAPIDS AI without any additional processing.
B) Use cuDF DataFrame operations to create new synthetic samples by applying random transformations (e.g., noise injection, permutation) to the existing dataset.
C) Use the RAPIDS cuML library to directly generate synthetic tabular data with controlled statistical properties.
D) Use numpy and pandas to generate synthetic data, then convert the DataFrame to cuDF before using it in RAPIDS AI workflows.

4. A financial institution is training a fraud detection model on a GPU-powered NVIDIA RAPIDS cuML pipeline. Their dataset includes customer ages, transaction amounts, merchant names, and transaction timestamps.
To optimize GPU memory usage while preserving accuracy, how should they store these features?

A) Keep customer ages in float16 format to reduce memory consumption, as integer types are less efficient in GPU operations.
B) Convert all numeric data to float64 for maximum precision, as rounding errors in lower precision could impact the model's accuracy.
C) Convert timestamps into UNIX epoch integers instead of using datetime64 for more efficient GPU computations.
D) Store transaction amounts as float32, customer ages as int8, merchant names as categorical, and timestamps as datetime64.

5. You are working with a GPU-based cloud environment and need to optimize the memory usage for a dataset that contains a column item_id representing unique product IDs. The item_id values are large integers, and there are over 10 million distinct product IDs.
Which of the following is the most memory-efficient data type choice for this column?

A) df['item_id'] = df['item_id'].astype('string')
B) df['item_id'] = df['item_id'].astype('int64')
C) df['item_id'] = df['item_id'].astype('int32')
D) df['item_id'] = df['item_id'].astype('float64')

Solutions: