# Roman Dubtsov

> Principal Engineer at NVIDIA

Location: Santa Clara, California, United States
Profile: https://flows.cv/romandubtsov

Looking for machine/deep learning, HPC or systems programming opportunities. Expertise: low-level tuning for x86, optimisations for manycore and cluster architectures, library engineering, API design, portability across OS/archs, heterogeneous programming.

I enjoy coding (have an undernourished open source project at http://stalonetray.sf.net), reading, and music.

## Work Experience
### Principal Engineer @ NVIDIA
Jan 2020 – Present | Santa Clara, California, United States
CUDA Math Libraries

### Principal Software Engineer @ Intel Corporation
Jan 2017 – Jan 2020 | Hillsboro, Oregon
Relocated. Leading development of MKL-DNN open-source deep learning library (https://github.com/01org/mkl-dnn).

### Principal Software Engineer @ Intel Corporation
Jan 2017 – Jan 2017 | Novosibirsk Region, Russian Federation
New title, same job. Leading development of MKL-DNN open-source deep learning library (https://github.com/01org/mkl-dnn).

### Sr. Software Engineer @ Intel Corporation
Jan 2015 – Jan 2017 | Novosibirsk Region, Russian Federation
Team lead. API design and performance optimization of popular compute operations used in DNN training and inference including cluster versions. Coordination across multiple HW and SW teams. Competitive positioning, performance projections, SW and HW pathfinding.

### Sr. Software Engineer @ Intel Corporation
Jan 2010 – Jan 2015 | Novosibirsk Region, Russian Federation
Led MKL BLAS Xeon Phi-related development in Russia. Set goals and objectives. Coordinated a number of cross-project and cross-Intel tools deliverables.

Launched a number of MKL versions for experimental architectures (key project coordinator at the early stages and one of the key developers later) and worked on productizations.

Worked on Xeon Phi DGEMM performance pathfinding from early stages to silicon debug.

### Sr. Software Engineer @ Intel Corporation
Jan 2006 – Jan 2010 | Novosibirsk
Member of a customer response team (CRT) focused on weather simulation benchmarks (mostly WRF). Established close relations with the developers community by helping with pre-release testing and making sure that all Intel tools are properly supported. Actively participated in related workshops. Published several posters and articles on improving WRF performance on Intel platforms. Published benchmarking results.


## Education
### PhD in Computer Science
Institute of Informatics Systems

### Master of Science in Applied Mathematics
Novosibirsk State University (NSU)


## Contact & Social
- LinkedIn: https://linkedin.com/in/roman-dubtsov-22570718

---
Source: https://flows.cv/romandubtsov
JSON Resume: https://flows.cv/romandubtsov/resume.json
Last updated: 2026-04-12