Abhishek Udupa
Software Engineer at Apple
Redmond, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolAbhishek Udupa is a software engineer with a decade of experience building high-performance systems and research-driven tooling, currently at Apple in Redmond. He has deep expertise in low-level GPU kernel engineering and end-to-end model optimization from his recent role as Principal Software Engineer at Microsoft, where he single-handedly optimized state-of-the-art LLMs to run on new Azure hardware. His background spans distributed systems, program synthesis, and formal methods from a PhD at the University of Pennsylvania, and earlier work includes building a safe network experimentation platform on the Azure WAN. An active contributor to the widely used ONNX Runtime project, he improved profiling, made ROCm/CUDA profilers session-aware and thread-safe, and enabled shape-sensitive GPU analysis to expose subtle performance bottlenecks. Comfortable reshaping vendor APIs and shipping production-grade optimizations, he blends rigorous academic training with pragmatic engineering that extracts real-world performance from emerging hardware. Notably, he pairs deep systems intuition with practical tooling improvements that accelerate diagnosing and tuning ML workloads.
10 years of coding experience
11 years of employment as a software developer
M.Sc (Engg.), Computer Science, M.Sc (Engg.), Computer Science at Indian Institute of Science (IISc)B.E, Computer Science, B.E, Computer Science at Manipal Institute of TechnologyDoctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at University of Pennsylvania
Github Skills (17)
performance-monitor10
performance-analytics10
gpgpu10
python10
onnx10
performance-measurement10
performance-analysis10
gpu10
profiling10
performance-tuning10
performance-monitoring10
hardware-acceleration9
roc9
cuda9
deep-learning8
Programming languages (4)
Github contributions (5)
microsoft/onnxruntime
Oct 2022 - Jan 2023
ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
Role in this project:
Performance Engineer
Performance EngineerContributions:140 reviews, 8 commits, 19 PRs in 2 months
Contributions summary:Abhishek focused on improving the performance and analysis capabilities of the ONNX Runtime. They developed a Python script for detailed profile analysis, enabling quicker identification of performance bottlenecks within the CPU and GPU kernels. Contributions also included enhancing the ROCm and CUDA profilers, making them session-aware and thread-safe, and fixing timestamp skews between CPU and GPU events. They also included algorithm selection exposed by ROCBLAS extensions API in GEMM autotuning and enabling shape-sensitive analysis in ProfileExplorer for GPU kernels.
runtimetrainingtensorflowai-frameworkaccelerator
abudup/libbpf
Jan 2020 - Mar 2022
Automated upstream mirror for libbpf stand-alone build.
Contributions:3 pushes in 2 years 1 month
stand-alonelibbpfstandupstream
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.