About

Divye Gala is a Senior Software Engineer at NVIDIA with a focus on building software that operates in a High-Performance Compute and/or Distributed environment. He is interested in scalable systems that have the capability to perform large-scale Machine Learning and Graph Analytics, as well as exploring real-time Inferencing solutions. With over 3 years of experience, Divye has contributed to several projects at NVIDIA. He has implemented Minimum Spanning Trees (MST) on the GPU using Boruvka’s algorithm, resulting in the quickest implementation with 5x to 48x speedups on random RMAT graphs found by comparing existing literature on the same architecture. He has also contributed to the GPU implementation of scikit-learn’s pairwise distance metrics for sparse matrices with comparable performance to cuSparse for inner product metrics, and no other implementations exist for other metrics. Divye has also helped execute the design and implementation of rapidsai/RAFT, a C++ template library for analytics primitives, setting up the public API, contribution guidelines, build system (CMake), and CI (docker). Additionally, he has created single and multi-GPU (using dask) dataset generators for blobs, classification, and regression to enable on-GPU memory testing frameworks. In his past roles at NVIDIA, Divye has worked as a Software Engineer and a Data Science Intern. He has ported implementation of HoltWinters algorithm to rapidsai/cuML, written unit tests, cython wrappers, and benchmarks, and improved python unit tests of rapids/cuML using fixtures to consolidate tests. He has also done exploratory work with xgboost+dask to run multi-GPU model training and inference. Divye holds a Master of Science in Computer Science from Georgia Institute of Technology and a Bachelor of Engineering in Computer Engineering from Bhartiya Vidya Bhavans. His tech stack includes Software Engineering, Scikit, test, Docker, ML, Data Science, XGBoost, and Python.