Albert Zeyer
Postdoctoral Researcher at RWTH Aachen University, Human Language Technology and Pattern Recognition Group
Aachen, North Rhine-Westphalia, Germany
Summary
👤
Senior🎓
Top SchoolAlbert Zeyer is a postdoctoral researcher and seasoned ML engineer with 17 years of experience bridging deep learning research and production-grade systems, particularly in speech recognition and sequence models. He holds a magna cum laude doctorate in computer science from RWTH Aachen and has driven both academic research and industry product development as Chief Scientist at AppTek while remaining active at RWTH. His open-source contributions include core CTC prefix score fixes and decoder improvements to the widely used ESPnet toolkit and low-level stability and data-handling work on the RWTH RETURNn framework, reflecting a focus on robustness and edge-case correctness. Comfortable in C++ and Python ecosystems, he also has practical systems experience (audio stacks, V8/JavaScript) and a long-standing habit of building side projects since childhood. Known for tackling hard theoretical and mathematical problems, he combines a deep theoretical background (automata, logic) with hands-on engineering that improves both research prototypes and deployed systems.
18 years of coding experience
1 year of employment as a software developer
Master of Science (MS), Computer Science, 1.6, Master of Science (MS), Computer Science, 1.6 at RWTH Aachen University
German, English, French
Stackoverflow
66,684reputation
6.6mreached
353answers
447questions
compression
top-5%
tcp
top-5%
firefox
top-5%
javascript
top-5%
subprocess
top-5%
cocoa
top-5%
Github Skills (58)
pytorch10
machine-translation10
udp10
python10
net10
airplay10
recurrent-neural-networks10
audio-processing10
c1110
portaudio10
networking10
c1710
deeplearning-ai10
deep-learning10
tensorflow10
Programming languages (20)
Github contributions (5)
rwth-i6/returnn
Jan 2015 - Dec 2022
The RWTH extensible training framework for universal recurrent neural networks
Role in this project:
Back-end Developer
Back-end DeveloperContributions:1797 reviews, 6414 commits, 1262 PRs in 8 years
Contributions summary:Albert's commits primarily focused on enhancing the RWTH extensible training framework for recurrent neural networks. Their contributions involved implementing data handling functionalities, specifically addressing dynamic data size and sequence length issues, along with tiling and beam fixes. They also worked on incorporating output dimensions and checking various performance aspects such as the CUDA error status. These changes indicate a focus on improving the framework's underlying computational mechanics and stability.
recurrent-neural-networksgputensorflowtheanodeep-learning
abrasive/shairport
Apr 2011 - Mar 2013
Airtunes emulator! Shairport is no longer maintained.
Role in this project:Back-end Developer
Back-end DeveloperContributions:108 commits in 1 year 11 months
Contributions summary:Albert primarily focused on enhancing the functionality of the shairport application, an Airtunes emulator. Their commits involved adding PortAudio support, refactoring the audio output mechanism by introducing options for libao and pipe output, and fixing endian issues. Moreover, they made improvements to the Squeezebox server integration for enhanced user experience. They also added features for a named pipe output.
emulatorshairport
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.