Saleel Kudchadker
Principal Member Of Technical Staff at AMD
Sunnyvale, California, United States
Summary
🤩
Rockstar🎓
Top SchoolSaleel Kudchadker is a Principal Member of Technical Staff at AMD with 8+ years focused on GPU compute runtimes, low-level performance engineering, and ML/LLM bottleneck analysis. He specializes in CUDA/HIP graph optimizations, micro-benchmarking, and runtime-backed architecture, driving performance and stability improvements across ROCm/HIP including VDI fixes, async copy handling, and elapsed-time accuracy. His tenure spans deep work on OpenCL drivers, GPU pipeline and AM model support, and systems-level features like SVM, WDDM 2.0, and IOMMUv2 for compute. Based in Sunnyvale, he pairs hands-on kernel and runtime tuning with production-facing problem solving, and his open-source contributions to the hip/ROCm ecosystem reflect practical impact on heterogeneous compute portability. He combines academic grounding in high-performance computing with a track record of resolving customer-critical performance bugs and enabling higher-level ML workloads.
8 years of coding experience
1 year of employment as a software developer
MS Computer Science, MS Computer Science at Arizona State UniversityBachelor of Engineering Electronics and Communication, Bachelor of Engineering Electronics and Communication at Goa University
Github Skills (15)
hip10
c-language10
cprogramming-language10
performance-monitoring9
async9
performance-measurement9
opencl9
performance-analytics9
performance-monitor9
performance-analysis9
asynchronous9
performance-tuning9
memory-management9
debug8
debugging8
Programming languages (4)
Github contributions (5)
ROCm/hip
Jul 2018 - Feb 2022
HIP: C++ Heterogeneous-Compute Interface for Portability
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:3 reviews, 60 commits, 11 PRs in 3 years 7 months
Contributions summary:Saleel primarily contributed to bug fixes and performance improvements within the HIP (Heterogeneous-Compute Interface for Portability) project, specifically focusing on the VDI (Virtual Device Interface) implementation. Their work involved resolving memory leaks, addressing issues with asynchronous memory copies and event handling. They also improved the accuracy of elapsed time calculations within the ROCclr runtime, and implemented features related to kernel configuration.
cudaheterogeneousgpuportabilityhip-runtime
saleelk/OpenCL-ICD-Loader
May 2018 - Jun 2019
The OpenCL ICD Loader project.
Contributions:14 pushes, 1 branch in 1 year 1 month
icdloaderopencl
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.