NVIDIA Corporation is an American technology giant based in Santa Clara, California. NVIDIA designs graphics processing units (GPUs) for the expanding gaming and professional markets, as well as system on chip units (SoCs) for the mobile computing and automotive market. Data scientists at NVIDIA leverage the most advanced data science technology in their products that ultimately get used in professional visualization, data centers, artificial intelligence, virtual reality, and self-driving cars.

Aside being a data-driven company, NVIDIA also helps data scientists across a wide range of industry (medicine, finance, operations etc.) make better use of their data and enhanced performance. With products like RAPIDS (an NVIDIA open source end-to-end GPU accelerated data science library that runs on Nvidia GPUs) and DGX line of GPUs, Nvidia has provided data scientists, machine learning/deep learning scientists and developers with the compute power they need to get their works done. [1]

The Data Science Role at NVIDIA

The roles of a data scientist at NVIDIA varies across specific teams, products, and features. Generally data scientist functions and roles at NVIDIA may span across a wide scope of data science concepts but are primarily focused on machine learning and deep learning. This means having an in-depth understanding of developing solutions on cloud computing clusters while also deploying ML and DL models at scale.

Required Skills

  • MS/PhD in Computer Science, Data Science, Electrical/Computer Engineering, Physics, Mathematics, other Engineering fields.
  • 3 years plus (8+ years for senior-level) work or research experience with Python, C++ software development.
  • Extensive experience with Machine Learning/Deep Learning algorithms with frameworks such as TensorFlow, PyTorch, XGboost, Scikit-learn, or Spark.
  • Proficient in the use of any of the following C, Python, Scala, SQL, Java, or C++ programming languages.
  • Ability to build ETL pipelines in a cloud environment.
  • Ability to cohesively work with multiple levels and teams across organizations (Engineering, Product, Sales and Marketing team).
  • Effective verbal/written communication, and technical presentation skills

What kind of data science role?

NVIDIA has no dedicated data science department, but there are a wide variety of data science teams each having its own unique process. There is a data science team working on data centers, a data team on RAPIDS, and an AI-driven auto team and software development team. Data scientist roles at NVIDIA may differ slightly and in some cases overlap.

Depending on the teams, the functions of a data scientist may stretch across being a deep learning engineer to primarily working as a research scientist focused on computer vision.

Basic responsibilities include:

  • Develop and demonstrate solutions based on NVIDIA’s state-of-the-art ML/DL, data science software and hardware technologies to customers
  • Perform in-depth analysis and optimization to ensure the best performance on GPU architecture systems.
  • Applying deep learning solutions to areas such as object detection, segmentation, video understanding, sequence prediction, adaptive computing, memory networks, reduced precision training and inference, graph compilers, reinforcement learning, search, distributed and federated training, and more.
  • Collaborate with key industry partner/customer developers to provide ML solutions applied to their products and technologies.
  • Partner with Engineering, Product and Sales teams to secure design wins at customers.
  • Work closely with customer's data science, ML/DL developers and IT teams.

The NVIDIA Interview Process

Deep learning developer guides

Like most tech companies, the data science hiring process at NVIDIA starts with an initial phone screen with recruiting, and then is followed by a technical phone screen with a team manager. After finishing through the technical phone screen you then proceed to the onsite interview comprised of 7 one-on-one interviews (each lasting between 30 to 60 minutes) with a hiring manager, team members, and a product manager.

Initial Screen

This is a resume-based phone interview with HR or a hiring manager. This interview is exploratory in nature and requires a run-down of your resume and relevant past projects to determine if you are a good fit for the position/team.

Technical Screen

After the initial phone screen, a technical interview with a data scientist will be scheduled. This interview is between 45 and 60 minutes long, and it involves questions around a real-life NVIDIA problem. Expect to explain your machine learning experience in-depth and talk about how you might design a ML or DL system and scale the process.

Onsite Interview

The onsite interview is the last interview stage in the NVIDIA data scientist hiring process. The onsite interview process for a data scientist at Nvidia comprises of 7 interview rounds, either one-on-one or with a small panel of interviewers, consisting of team members, a team manager, and a product manager, with each interview rounds lasting between 45 and 60 minutes.

This interview is a combination of various data science concepts including data analytics, software engineering, machine learning, and NVIDIA’s core culture and values. For the technical questions, candidates are expected to perform coding exercise on a whiteboard or on a laptop provided by NVIDIA.

Questions in this interview span across advanced statistical concepts to deep learning implementation and design. For the technical aspect, remember to practice coding in compiler languages as well as Python, and also practice questions on machine learning algorithms and coding in Tensorflow, Keras, or other deep learning frameworks.

Last Tips

  • It really helps to have prior experience in GPU development. This means GPGPU programming and design practices or working for a potential competitor such as Intel, AMD, etc... in some sort of data science capacity.
  • NVIDIA has been breaking the technology barrier with their GPU and chip development. This generally done by hiring researchers or people out of academia which established records of thought leadership in a technical area or industry segment. This also means they are willing to pay an extraordinary sum in total compensation to do so as well.

NVIDIA Data Science Interview Questions

  • Given a time series dataset, how would you detect an anomaly?
  • What's the difference between a True Positive and a False Positive?
  • Implement gradient descent in Tensorflow.
  • Design a recommendation engine from end to end from a dataset to deployment in production.
  • Write down the equation for linear regression.
  • Explain how a decision tree works under the hood.

References

[1] Deep Learning AI