Theodoros Vasiloudis
Senior Applied Scientist at Amazon Web Services (AWS)
Seattle, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolTheodoros Vasiloudis is a Senior Applied Scientist with 11 years’ experience building scalable machine learning systems for industry leaders including Amazon, Spotify, Pandora and Data Artisans. He holds a PhD from KTH focused on scalable ML via approximation and distributed computing, and has practical expertise in large-scale graph ML, gradient boosted trees, online learning and LLM fine-tuning. At AWS he has driven distributed graph learning infrastructure (GraphStorm, DGL integrations) and now designs large-scale model customization and RL/feedback-driven fine-tuning for Bedrock and SageMaker. He is an active open-source contributor and Apache Flink committer, with concrete contributions to DGL that improved distributed graph partitioning and Parquet-based data handling. Colleagues describe him as a practitioner who closes the gap between cutting-edge research and production at multi-billion-edge scale. Based in Seattle, he combines deep academic training with a track record of speeding up model training and evaluation by orders of magnitude.
11 years of coding experience
11 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at KTH Royal Institute of TechnologyBachelor's degree, Software Engineering, Bachelor's degree, Software Engineering at Computer Science Department, Aristotle University of Thessaloniki
English, Greek
Github Skills (10)
parquet10
deep-learning10
graph-neural-network10
pyarrow10
python10
distributed-computing10
numpy9
pytorch8
data-pipeline8
data-pipelines8
Programming languages (16)
Github contributions (5)
dmlc/dgl
Jan 2023 - Jan 2023
Python package built to ease deep learning on graph, on top of existing DL frameworks.
Role in this project:
Backend Developer
Backend DeveloperContributions:27 reviews, 2 commits, 7 PRs in 2 days
Contributions summary:Theodoros primarily focused on enhancing the DGL library's data handling capabilities, particularly for graph partitioning. They implemented support for Parquet-formatted edge files and single-column vector Parquet files, improving data loading flexibility. Their contributions included modifications to the distributed partitioning tools, specifically `distpartitioning`, as well as adjustments to unit tests to accommodate the new file formats and functionalities. The user also addressed issues within data shuffling and the GraphBolt conversion process, demonstrating a focus on improving the library's efficiency and usability for distributed graph processing.
pytorchpythondeep-learningmachine-learninggraph-neural-networks
thvasilo/flink
Mar 2015 - Oct 2016
Mirror of Apache Flink
Contributions:105 pushes, 34 branches, 1 comment in 1 year 6 months
flinkapachebig-dataapache-flinkjava
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.