I am currently working as a Senior Compute Architect at NVIDIA.
My work focuses on implementation and optimization of Math and Deep Learning libraries such as CUTLASS and others.
I graduated with a PhD in Computer Science from University of California, Irvine (UCI).
My research interests lie in the fields of Compilers and High-Performance Computing.
More specifically, I have worked on Loop Optimizations (including Auto-Vectorization techniques and Polyhedral Model based techniques) as part of my PhD dissertation.
I have also worked on the use of Artificial Intelligence (AI) to assist compilers and compiler optimizations, and study the impact of compiler optimizations on energy efficiency.
I developed a meta-compilation framework, the MCompiler, for improved performance and efficiency.
Previously, I have interned at NVIDIA, Intel and Samsung Research America working on various compilers, runtime libraries and performing hardware-oriented optimizations for CPUs and GPUs.
A Multiple Compiler Framework for Improved Performance
A. Shivam, A. Nicolau, A.V. Veidenbaum
Languages and Compilers for Parallel Computing. LCPC 2023.
OpenACC Routine Directive Propagation using Interprocedural Analysis
A. Shivam, M. Wolfe
WACCPD 2018 - Fifth Workshop on Accelerator Programming Using Directives, SC 2018.
Towards an Achievable Performance for the Loop Nests
A. Shivam, N. Watkinson, A. Nicolau, D. Padua, A.V. Veidenbaum
Languages and Compilers for Parallel Computing. LCPC 2018.
Load Balancing with Polygonal Partitions
A. Shivam, P. Ravi, A.V. Veidenbaum, A. Nicolau and R. Cammarota
IMPACT 2018 - 8th International Workshop on Polyhedral Compilation Techniques, HiPEAC 2018.
Using Hardware Counters To Predict Vectorization
N. Watkinson, A. Shivam, Z. Chen, A.V. Veidenbaum, A. Nicolau, Z. Gong
Languages and Compilers for Parallel Computing. LCPC 2017.
Polygonal Iteration Space Partitioning
A. Shivam, A. Nicolau, A.V. Veidenbaum, M.M. Furnari, R. Cammarota
Languages and Compilers for Parallel Computing. LCPC 2016.
Email: aniketsh AT uci DOT edu