Doru Thom Popovici is a Research Scientist at Lawrence Berkeley National Laboratory, working at the intersection of algorithms, compilers, and hardware. His research focuses on the development of frameworks and methodologies that enable scientific applications to achieve high performance and portability across current and emerging hardware platforms, spanning CPUs, GPUs, and novel architectures.

His work covers a broad range of topics in high performance computing, including domain-specific languages, performance modeling, parallel computing, and code generation. He has made significant contributions to the design and optimization of Fourier transforms and linear algebra operations, with applications in density functional theory, quantum circuit simulation, and graph analytics. He is a key contributor to the SPIRAL system for extreme performance portability and has led work on flexible, multi-dimensional FFT frameworks, bandwidth-efficient FFT algorithms, and automatic generation of mappings for distributed Fourier operations.

More recently, his research has expanded into quantum computing, where he develops graph partitioning techniques for quantum circuit simulation, and into machine learning, where he has co-authored work on memory-efficient fine-tuning of transformer-based models. He has also contributed to hardware design space exploration methodologies for evaluating future HPC architectures.

Popovici earned his Ph.D. from Carnegie Mellon University under the supervision of Professor Franz Franchetti.