OpenExamsFREE PRACTICE — ANY EXAM

NVIDIA Accelerated Data Science certification exam — Information & Free Practice

100 free questions · Answer explanations · No signup required

Start free practice →

Exam Review

About NVIDIA Accelerated Data Science certification exam

NVIDIA Accelerated Data Science certification exam · US

Key exam facts

Provider
NVIDIA; registration through the NVIDIA-linked Certiverse portal
Format
English, online and remotely proctored, question-based pass/fail examination; numerical score and item-type shares are not published in the reviewed materials
Exam questions
Total questions: 60–70.
Time limit
120 minutes
Exam fee
USD 200

Topics covered

  • Data Analysis (14%)

    • Detect anomalies in a time-series dataset
    • Conduct time-series analysis
    • Create and analyze graph data using something like cuGraph
    • Identify how much data is big data (or when to use which acceleration method)
    • Perform exploratory data analysis (EDA)
    • Visualize time-series data

  • Data Manipulation and Software Literacy (19%)

    • Design and implement ETL workflows using accelerated ETL processes
    • Implement data caching to reduce shuffle
    • Use distributed data processing frameworks to process big data
    • Implement data parallelism using Dask for multi-GPU scaling
    • Profile deep learning models using tools such as DLProf
    • Determine the optimal data processing libraries to use for varying dataset sizes
    • Transform and standardize data using cuDF
    • Manage and allocate GPU memory effectively to improve performance, utilizing strategies to optimize memory
    • Scale inference for large datasets in both single-node and multi-node systems

  • Data Preparation (17%)

    • Perform data cleansing and preprocessing using cuDF and pandas
    • Transform and standardize data
    • Standardize data as needed to ensure uniformity across features
    • Identify and acquire datasets
    • Monitor data processing pipelines to recognize bottlenecks
    • Process, organize, and store datasets
    • Generate synthetic data to augment datasets using cuDF and NVIDIA RAPIDS

  • GPU and Cloud Computing (16%)

    • Analyze graph data using GPU-accelerated tools like cuGraph
    • Optimize performance of the data science process through GPU acceleration
    • Describe, follow, and execute the CRISP-DM process
    • Utilize dependency management frameworks, such as Docker and Conda to manage software-versioning conflicts
    • Determine the optimal data type choice for each feature
    • Compare frameworks' performance by designing and implementing a benchmark

  • Machine Learning (15%)

    • Split data, choosing the best split type and method
    • Standardize data as needed to ensure uniformity across features
    • Identify how much data is big data (or when to use which acceleration method)
    • Perform rapid experimentation to find the balance between model accuracy and inference performance
    • Optimize hyperparameters of machine learning models
    • Train machine learning models for both single-GPU and multi-GPU scenarios
    • Use GPU memory optimization techniques, such as batching and mixed precision, to train machine learning models
    • Profile deep learning models using tools such as DLProf

  • MLOps (19%)

    • Determine the optimal data type choice for each feature
    • Assess and verify the size in memory of a dataset
    • Compare the required memory with the available memory on a device
    • Perform benchmarking and optimize different GPU-accelerated workflows
    • Deploy and monitor models in production

How to prepare

Work through the topics, read the explanation after each answer, and revisit the areas where you miss questions. Use the official outline and reference materials below to guide your study.

These independent practice questions are not actual exam items and do not guarantee a passing result.

Frequently asked questions

What is the exam format?

English, online and remotely proctored, question-based pass/fail examination; numerical score and item-type shares are not published in the reviewed materials

How many questions are on the exam?

Total questions: 60–70.

How long is the exam?

120 minutes

How much does the exam cost?

USD 200

Official sources and references

Confirm current exam arrangements with the exam provider.