Ivan Stepanov
Machine Learning Engineer at Upwork
Serbia
Summary
🤩
Rockstar🎓
Top SchoolIvan Stepanov is a Machine Learning Engineer with 10 years of hands-on experience building high-performance systems and shipping production ML solutions from edge devices to GPU clusters. Based in Serbia, he has optimized models for constrained environments (YOLOv5 on Android/TFLite), ported compute-heavy workloads to CUDA with >10,000× speedups, and architected a GPU cloud marketplace and vector search pipelines for 100M+ messages. He contributes to popular open-source LLM tooling (llama.cpp), improving sampling algorithms and cross-platform robustness, and has deep C++/Python backend experience from medical systems to embedded gaming hardware. Known for pragmatic performance engineering, Ivan also mentors engineers and has a track record of turning research ideas into deployable, resource-efficient products.
10 years of coding experience
4 years of employment as a software developer
Bachelor's degree, Bachelor's degree at Moscow Technical University of Communications and Informatics (MTUCI)
Stackoverflow
1reputation
0reached
1answer
0questions
Github Skills (18)
algorithm10
algorithms10
c-language10
subsampling10
llama10
sampling10
llm10
gm10
sampler10
upsampling10
gml10
cprogramming-language10
datatypes7
windows7
data-handling7
Programming languages (24)
Github contributions (5)
ggml-org/llama.cpp
Apr 2023 - Nov 2024
LLM inference in C/C++
Role in this project:
Back-end Developer
Back-end DeveloperContributions:27 reviews, 16 PRs, 4 pushes in 1 year 7 months
Contributions summary:Ivan's contributions primarily focus on enhancing the llama.cpp project's sampling algorithms and handling of edge cases. They introduced new sampling techniques such as locally typical, tail-free, and frequency/presence penalties, improving the model's generation capabilities. Furthermore, the user made improvements to handle temperature sampling and incorporated bug fixes related to Windows builds and data type support (BF16). They also made adjustments to exception handling within the core C++ library.
ggmlllama
ivanstepanovftw/osu-FakinKrakin
Jun 2018 - Apr 2022
Contributions:1 release, 1 tag in 3 years 10 months
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.