Leonard Lausen
Applied Scientist at Amazon Web Services
New York, New York, United States
Summary
🤩
RockstarLeonard Lausen is an Applied Scientist at Amazon AI with 14 years of experience building and optimizing ML systems, distributed training, and build/CI infrastructure. He has hands-on expertise improving deep learning frameworks—contributing performance and optimizer fixes to Apache MXNet, integrating MXNet with Horovod for distributed training, and enhancing NLP tooling in gluon-nlp. Leonard pairs back-end ML engineering with devops and build automation skills, having modernized CI, cross-platform builds, and Docker-based environments for large open-source projects. He also contributes to developer tooling and editor ecosystems, showing an attention to developer experience beyond core ML stacks. Based in New York, he blends production-focused engineering with research-minded rigor, often addressing subtle issues like optimizer weight decay handling and memory-leak fixes that materially improve runtime stability.
14 years of coding experience
Stackoverflow
532reputation
14kreached
9answers
4questions
Github Skills (43)
tokenize10
github-ci10
c-language10
distributed-training10
python10
package-management10
configuration-management10
mxnet10
machine-learning10
data-serialization10
cmake10
horovod10
cicd10
fasttext10
serialization10
Programming languages (27)
Github contributions (5)
dmlc/gluon-nlp
Apr 2018 - Apr 2021
NLP made easy
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:3 releases, 76 reviews, 228 commits in 3 years
Contributions summary:Leonard primarily contributed to the development of the gluon-nlp library by fixing serialization issues in the Vocab and TokenEmbedding classes. The user also added support for loading and evaluating pretrained fastText embeddings in different file formats, further improving the utility of the library for NLP tasks. In addition, the user modified the dataset to support the new data format in the system. The contributions encompass both back-end and ML engineering.
nlpmxnetnatural-language-understandingnlgnatural-language-generation
apache/mxnet
Dec 2016 - Jun 2021
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:7 releases, 261 reviews, 266 commits in 4 years 7 months
Contributions summary:Leonard contributed to the Apache MXNet project, focusing on improving the performance and functionality of optimization algorithms and deep learning operators. They addressed issues related to the Adam and RMSProp optimizers, including handling weight decay, implementing weight clipping, and introducing uncentered RMSProp. The user also implemented 3D deconvolution for cuDNN and made adjustments to existing code to support it. Further contributions include enhancements to the Python API for serialization and the handling of memory and memory leaks.
pythonschedulerdataflowmutationdata-science
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.