About

You can read more about my education & work experience below.

Svetlozar Georgiev

Software Engineer

Summary

Software engineer with 6 years of industry experience, specialising in low-level programming and modern C++. Strong background in systems development, performance-critical code, and cross-platform software engineering. Skilled in working close to the metal, with a solid understanding of memory management, concurrency, and hardware-aware optimisation. Experienced in GPU programming (including SYCL and CUDA) and familiar with graphics and compute APIs such as Vulkan and OpenGL. Brings a rigorous, detail-oriented approach to solving complex technical problems in high-performance environments.

Experience

Senior Software Engineer — GPU-Accelerated Machine Learning Libraries
Codeplay Software (Intel Subsidiary) · Edinburgh
Sep 2019 – Present

Responsibilities

• Designed and implemented complex software features using modern C++17/20, focusing on performance and maintainability.
• Developed cross-platform software targeting Linux and Windows, using CMake for build configuration.
• Collaborated in cross-functional Agile teams, planning features in Jira and reviewing code in structured PR workflows.
• Led projects as Product Owner, coordinating with stakeholders, prioritising features, and defining technical roadmaps.
• Delivered reliable, production-ready code integrated into open-source and proprietary software and libraries.

Key projects

• oneDNN NVIDIA Backend: Integrated cuDNN and cuBLAS by mapping oneDNN operations to vendor-optimised implementations.
• Binary DNN Model Loaders: Extended ONNX, PyTorch, and Caffe2 model importers to deploy models to custom accelerator hardware, utilising a proprietary graph-lowering compiler for optimisations.
• SYCL-DNN Library: As Product Owner, led the development of portable, high-performance SYCL GPU kernels for Codeplay’s deep-learning library.
• ONNX Runtime SYCL Backend: Developed and integrated a SYCL backend for ONNX Runtime operators, using SYCL-DNN and custom kernels developed by the team.
• oneDNN SYCL Backend: Led a team working alongside Intel’s core oneDNN team to expand SYCL kernel support and performance for multiple GPU vendors.
• CI/CD + Benchmarking: Built automated test and benchmark pipelines using Grafana for visualisation.
• llama.cpp SYCL Backend: Maintaining and optimising the SYCL backend of llama.cpp for performance, portability, and correctness.

Skills

GPU APIs

SYCL CUDA CUDA Libraries OpenGL Vulkan

Languages

C C++ Python Bash GLSL

Tools

Git CMake Vtune NVIDIA Nsight RenderDoc Docker

Machine Learning Frameworks

PyTorch ONNX Runtime Caffe2 TensorFlow

Education

Bachelor of Engineering with Honours (1st Class)
Edinburgh Napier University · Edinburgh
Sep 2015 – May 2019

• Modules included: Software Engineering Methods, Computer Graphics, Database Systems, Mathematics, Computer Systems, Concurrent & Parallel Systems, Games Engineering, Algorithms & Data Structures, Data Analytics.
• Dissertation: Application of Chatbots in Education — Developed a web-based chatbot with responsive design using Python (Flask, Chatterbot) and Bootstrap, capable of answering student FAQs via NLP trained on Stack Overflow data.

Publications

• Arroyo, H., Keir, P., Angus, D., Matalonga, S., Georgiev, S., Goli, M., Dooly, G. & Riordan, J. (2024). Segmentation of drone collision hazards in airborne RADAR point clouds using PointNet. IEEE Transactions on Intelligent Transportation Systems, 25(11), 17762–17777. https://doi.org/10.1109/TITS.2024.3442668
• Angus, D., Georgiev, S., Arroyo Gonzalez, H., Riordan, J., Keir, P. & Goli, M. (2023). Porting SYCL accelerated neural network frameworks to edge devices. In: Proceedings of the 2023 International Workshop on OpenCL (IWOCL ’23), Cambridge, UK. Article 4, 1 p. https://doi.org/10.1145/3585341.3585346
• Narasimhan, K., El Farouki, O., Goli, M., Tanvir, M., Georgiev, S. & Ault, I. (2022). Towards performance portability of AI graphs using SYCL. In: Proceedings of the IEEE/ACM International Workshop on Performance, Portability and Productivity in HPC (P3HPC), 2022, pp. 111–122. https://doi.org/10.1109/P3HPC56579.2022.00016
• Tanvir, M., Narasimhan, K., Goli, M., El Farouki, O., Georgiev, S. & Ault, I. (2022). Towards performance portability of AI models using SYCL-DNN. In: Proceedings of the 10th International Workshop on OpenCL (IWOCL ’22), Bristol, UK. Article 23, 3 pp. https://doi.org/10.1145/3529538.3529999
• Goli, M., Narasimhan, K., Reyes, R., Tracy, B., Soutar, D., Georgiev, S., Fomenko, E. M. & Chereshnev, E. (2020). Towards cross-platform performance portability of DNN models using SYCL. In: 2020 IEEE/ACM International Workshop on Performance, Portability and Productivity in HPC (P3HPC), 2020, pp. 25–35. https://doi.org/10.1109/P3HPC51967.2020.00008