Marko Radmilac
Software Engineer at Reflection AI
Bellevue, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolMarko Radmilac is a seasoned software engineer with 11 years focused on building and optimizing large-scale distributed systems and AI infrastructure, currently working on RL and agentic infrastructure. He led platform and GPU cluster efforts at Microsoft—helping train high-profile models like Microsoft Phi-3 and supporting OpenAI deployments on AzureML—and briefly contributed to AI infrastructure at Meta. His early work spans deep performance engineering from C++ compiler backends to distributed build caching and big-data runtimes, and he has hands-on ML toolkit experience contributing CUDA kernel and prefetching optimizations to the Microsoft CNTK project. Based in Bellevue, WA, Marko combines applied mathematics training (M.S., University of Washington) with decades of low-level optimization expertise to squeeze efficiency from both hardware and software. He’s comfortable moving between algorithmic performance tweaks and production-scale orchestration, with a track record of turning research-scale training needs into reliable, production-grade systems.
10 years of coding experience
1 year of employment as a software developer
Master's degree, Applied Mathematics, Master's degree, Applied Mathematics at University of WashingtonNone (PhD), Computer Science, None (PhD), Computer Science at Brown UniversityBachelors, Computer Science, Applied Math, Bachelors, Computer Science, Applied Math at Florida Institute of Technology
Stackoverflow
11reputation
67reached
1answer
0questions
Github Skills (16)
neural-network10
cuda10
gpu-programming10
machine-learning10
c-language10
cntk10
deep-learning10
cprogramming-language10
optimization10
cplus9
cpp9
build-automation8
python8
speech6
azure-cognitive-services6
Programming languages (8)
Github contributions (5)
microsoft/CNTK
Jun 2015 - Sep 2016
Microsoft Cognitive Toolkit (CNTK), an open source deep-learning toolkit
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:62 commits, 20 pushes, 7 branches in 1 year 4 months
Contributions summary:Marko contributed to the Microsoft Cognitive Toolkit (CNTK) by modifying CUDA kernels and build scripts. Their work focused on optimizing performance, including fixing thread divergence issues in sigmoid kernels and optimizing code for zero value settings. They also implemented a minibatch prefetching mechanism for data reading optimization. Further contributions included adjustments to the build process and making sure Linux and Windows builds used consistent integer types within CUDA kernels.
pytorchpythondeep-learningc-plus-plusmachine-learning
mradmila/unimrcp
Oct 2019 - Feb 2022
Open source cross-platform implementation of MRCP protocol
Contributions:2 PRs, 15 pushes, 3 comments in 2 years 4 months
mrcpcross-platform
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.