Vinay Deshpande
Senior DevTech Engineer at NVIDIA
Pune, Maharashtra, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolVinay Deshpande is a Senior DevTech Engineer with over a decade of experience specializing in parallel computing, GPU programming, performance optimization, and mathematical algorithm implementation using C, C++ and Python. Based in Pune and currently at NVIDIA, he has a long tenure contributing to high-performance ML infrastructure, notably improving RNGs and DBSCAN implementations in RAPIDS libraries like raft and cuML. His work blends low-level CUDA expertise with practical engineering—optimizing device memory, CUDA streams, and resource handles to make algorithms production-ready. Earlier roles at IBM and Computational Research Lab built his foundations in systems and instrumentation, supported by an M.Tech from IISc. He’s known for tackling subtle reliability issues (e.g., RNG correctness and sample-without-replacement edge cases) that often go unnoticed but are critical for ML reproducibility.
10 years of coding experience
17 years of employment as a software developer
M. Tech, Instrumentation, M. Tech, Instrumentation at Indian Institute of Science (IISc)B. E., Instrumentation, B. E., Instrumentation at Swami Ramanand Teerth Marathwada University
Github Skills (21)
algorithm10
algorithms10
c-language10
gpu-programming10
machine-learning10
gpu10
dbscan10
cuda10
randomization10
cprogramming-language10
testing9
numerical9
machine-learning-algorithms9
numerical-methods9
numeric9
Programming languages (7)
Github contributions (5)
rapidsai/cuml
Mar 2019 - Aug 2022
cuML - RAPIDS Machine Learning Library
Role in this project:
ML Engineer
ML EngineerContributions:75 reviews, 131 commits, 41 PRs in 3 years 5 months
Contributions summary:Vinay focused on implementing and testing the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm within the cuML library. Their contributions involved adding a DBSCAN example, patching existing test cases, and integrating new code to improve performance. The commits also demonstrate the use of CUDA streams for asynchronous operations and memory management, including device allocation and deallocation. The user updated the code to utilize a `cumlHandle` for managing CUDA resources.
cudacumlnvidiadata-sciencegpu
rapidsai/raft
Feb 2022 - Dec 2022
RAFT contains fundamental widely-used algorithms and primitives for machine learning and information retrieval. The algorithms are CUDA-accelerated and form building blocks for more easily writing high performance applications.
Role in this project:ML Engineer
ML EngineerContributions:7 reviews, 7 commits, 8 PRs in 10 months
Contributions summary:Vinay primarily contributed to the RAFT library by improving and fixing the random number generator (RNG) functionality, which is crucial for many machine learning algorithms. Their work included removing inferior generators, optimizing existing ones, adding new generators like PCG, and providing device APIs for different probability distributions. They also addressed test failures related to the RNG and sample-without-replacement functions, indicating a focus on ensuring the reliability of these components for use in machine learning applications.
sciencedata-sciencemachine-learninggraph-data-sciencefundamental
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.