Node RAPIDS

Skills: C++, TypeScript, CUDA

During my internship at NVIDIA, I had the chance to work on Node RAPIDS. It is a collection of Node.js native addons that provide access to GPU-accelerated data science libraries. Essentially, it allows data scientists to skip learning CUDA and instead use the JavaScript interface to accelerate data science operations on the GPU. A significant portion of my contributions involved writing bindings to the cuDF library. This process included writing the TypeScript interface, translating the call into the C++ layer, and finally invoking the cuDF and CUDA layers.

Towards the latter half of my internship, I was tasked with setting up bindings for Blazing SQL. I have a blog post discussing the implementation details of the Blazing SQL library here. Essentially, we integrated a GPU-accelerated SQL engine with Node.js.