Prethvi Kashinkunti
Manager, Datacenter Systems Engineering
Greater Chicago Area United States
Summary
🤩
Rockstar🎓
Top SchoolPrethvi Kashinkunti is a Manager of Datacenter Systems Engineering at NVIDIA with eight years of experience designing and deploying AI and deep learning infrastructure for cloud providers and enterprise customers. He combines an academic foundation in HPC and parallel computer architecture from Northwestern with hands-on systems and solutions work across NVIDIA engineering roles. Prethvi has driven performance-focused ML engineering efforts—contributing to NVIDIA’s retinanet-examples with mixed-precision training, DALI dataloading, and ResNeXt backbone integrations—to squeeze GPU throughput and reliability in production workflows. He is fluent at bridging customer requirements and low-level system optimizations, turning research-grade models into scalable, validated datacenter solutions. Based in the Greater Chicago Area, he brings a blend of technical leadership, practical systems tuning, and an attention to operational readiness honed from early roles in research and utility forecasting.
8 years of coding experience
8 years of employment as a software developer
Anderson High SchoolMaster's Degree Computer Engineering, Master's Degree Computer Engineering at Northwestern University
Github Skills (9)
multiprecision10
retinanet10
cuda10
object-detection10
computer-vision10
pytorch10
resnet9
onnx9
deep-learning9
Programming languages (3)
Github contributions (5)
NVIDIA/retinanet-examples
Mar 2019 - Sep 2019
Fast and accurate object detection with end-to-end GPU optimization
Role in this project:
ML Engineer
ML EngineerContributions:1 release, 39 commits, 12 PRs in 6 months
Contributions summary:Prethvi's contributions primarily involve modifying and updating the `retinanet-examples` repository, focusing on object detection. The commits show a transition to a new AMP unified API, indicating work on mixed-precision training to improve performance. The changes include updates to the training, inference, and loss functions, alongside dataset improvements using DALI, and the addition of ResNeXt backbones. Further work involved fixing issues related to validation and improvements to the DALI dataloader.
pytorchend-to-enddeep-learninggpuoptimization
pkashinkunti/DALI
Dec 2018 - Jan 2019
A library containing both highly optimized building blocks and an execution engine for data pre-processing in deep learning applications
Contributions:9 pushes, 1 branch in 1 month
pytorchdata-pre-processingpre-processingbuilding-blocksdeep-learning
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.