Tucker Arrants

Bachelor of Science and Arts - United States

Tucker is a Physics and Philosophy graduate from Chapman University who is interested in computer vision and biological data science. He likes to spend his free time on Kaggle, where he earned his 109th ranking out of more than 138,000 data scientists. When he isn’t coding on Kaggle, he enjoys playing tennis and video games.

I was born not knowing and have had only a little time to change that here and there.

Richard Feynman

Theoretical physicist

  • What can you now do on a Z that you couldn’t do before?

    Never before have I been able to iterate through my experiments so quickly: I am able to go from a hypothesis to evaluating the results faster than ever, which allows me to test many more ideas.

  • What is the next thing you want to learn or experiment with on your Z?

    I want to learn more about vision transformers, and run my own experiments on the hardware for current and upcoming Kaggle computer vision projects.

  • How has access to powerful NVIDIA GPUs helped to push your work further?

    Having fast NVIDIA Quadro RTX graphics cards allows me to run GPU experiments quicker than before and having my own local GPU gives me more control over my projects. I don't have to worry about running out of Kaggle / Google Colab quota and can run long experiments that would usually timeout when using Cloud GPU services. Not only am I able to train models quicker, having enormous GPU memory gives me the ability to train with large batch sizes, which is important when you have noisy data (which you normally do in real world datasets). On top of that, it is common practice to decrease the learning rate during training, but this unfortuantely slows down the convergence speed of your model. That being said, there are several papers showing that you can obtain the same learning curves by instead increasing the batch size during training. So now that I have all this free GPU memory to play with, I can simply increase the batch size during training instead of decreasing the learning rate, allowing me to iterate even faster than ever.

  • What’s your favorite Z feature?

    The preloaded data science software on the ZBook. I know very well the struggle of installing all the compatible versions of various softwares to finally get your data science projects up and running. With the ZBook, you’re able to take it out of the box and get started without wasting time—that you could otherwise be using to run experiments.

My current gear provided by Z and Nvidia
  • Z8

    When you work with complex simulations, data analysis or high-end VFX, only the best will do. Cut through your toughest jobs with our flagship Z - built for power-intensive workloads.

    Learn more

    Never has so much power been engineered into such a sleek package. Keep your best talent creating on the latest tech with the new benchmark for mobile workstations.

    Learn more
  • Dual Z27

    Experience extraordinary, precise color and see more of your projects at once on a display with stunning 4K resolution.

    Learn more
  • Preloaded Software Stack

    Save countless days installing, configuring and maintaining your data science environment by choosing the preloaded data science software stack.

    Learn more
Ambassador Updates

Stay informed and get the latest Data Science news and updates

Stay in touch