Marek Travnikar
Staff Software Engineer - Calibration
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolMarek Travnikar is a Staff Software Engineer specializing in calibration with 12 years of experience building reliable, production-ready systems for autonomous vehicle platforms in the San Francisco Bay Area. At Aurora he progressed from individual contributor roles to staff level, focusing on calibration infrastructure that bridges sensor behaviour, machine learning models, and vehicle control. He brings practical ML engineering experience from contributing performance and optimizer improvements to the widely used Apache MXNet deep-learning framework, including efficiency gains to sequence operators. Marek combines a robotics engineering foundation from Worcester Polytechnic Institute with hands-on hardware verification internships and systems work at Uber and Cray, giving him fluency across software, hardware, and tooling. Known for pragmatic optimizations and shipping measurable improvements, he prefers solving ambiguous, cross-disciplinary problems that require both algorithmic and systems thinking.
12 years of coding experience
3 years of employment as a software developer
Bachelor of Science - Robotics Engineering, Bachelor of Science - Robotics Engineering at Worcester Polytechnic Institute
Github Skills (11)
mxnet10
lstm10
rnn-model10
deep-learning10
python10
n10
machine-learning9
optimization8
scala5
tensorflow5
cuda4
Programming languages (9)
Github contributions (5)
apache/mxnet
Feb 2017 - Aug 2018
Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Scala, Go, Javascript and more
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:5 commits, 7 PRs, 94 comments in 1 year 5 months
Contributions summary:Marek primarily contributed to the Apache MXNet deep learning framework by fixing bugs and implementing improvements to existing functionalities. Their work involved correcting dataset URLs for a character RNN example, improving the PTB bucketed LSTM model by adding dropout support and non-SGD optimizers. Furthermore, they optimized the sequence reverse operator to enhance the efficiency of sequence processing operations.
pythonschedulerdataflowmutationdata-science
mkolod/mxnet_seq2seq
Apr 2017 - Jun 2017
Contributions:244 commits, 3 PRs, 235 pushes in 2 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.