Iaroslav Shcherbatyi
Senior Machine Learning Engineer - Amazon Q at Amazon Web Services (AWS)
Berlin, Germany
Summary
🤩
Rockstar🎓
Top SchoolIaroslav Shcherbatyi is a Senior Machine Learning Engineer with a decade of experience building and shipping production-scale ML and GenAI systems, currently working on quality and security for Amazon Q at AWS in Berlin. He has a strong record in SageMaker and Autopilot—designing automated ML and generative capabilities—and maintains open-source work in hyperparameter optimization (scikit-optimize). His background spans research labs (Max Planck, DFKI, Universität des Saarlandes) where he bridged theory and practice in computer vision and healthcare applications, and he has contributed practical RL environments and language-model experiments to prominent projects like OpenAI Gym and Requests for Research. Known for turning research insights into reliable services, he combines rigorous experimental thinking with production engineering, including building convergence-control environments for training CNNs. Fluent in Python and tooling around ML workflows, he brings both academic depth and sizable cloud-scale product experience to generative AI reliability.
10 years of coding experience
8 years of employment as a software developer
Master's degree, Computer Science, Master's degree, Computer Science at Universität des Saarlandes
English, Russian, Ukrainian
Github Skills (22)
python10
scikit10
gymnasium10
openai-gym10
machine-learning10
reinforcement-learning10
mask-rcnn10
hyperparameter-optimization10
keras10
bayesian10
deep-learning10
optimisation10
language-modeling10
scikit-learn10
neural-network10
Programming languages (10)
Github contributions (5)
scikit-optimize/scikit-optimize
Apr 2017 - Apr 2019
Sequential model-based optimization with a `scipy.optimize` interface
Role in this project:
Data Scientist & ML Engineer
Data Scientist & ML EngineerContributions:5 releases, 96 commits, 57 PRs in 2 years
Contributions summary:Iaroslav contributed to the implementation of a supervised learning benchmark for black-box optimization algorithms. They focused on optimizing parameters for various machine learning models from scikit-learn, including SVR, SVC, DecisionTreeRegressor, DecisionTreeClassifier, MLPClassifier, and MLPRegressor. The changes include the addition of multiple surrogate models for optimization algorithms and updates to the benchmark code to improve functionality.
pythonbayesoptscipymodel-based-optimizationbayesian-optimization
openai/requests-for-research
Jun 2016 - Jun 2016
A living collection of deep learning problems
Role in this project:Data Scientist
Data ScientistContributions:11 commits, 3 PRs, 7 comments in 17 days
Contributions summary:Iaroslav contributed to a research request focused on building a language model for generating jokes. They iterated on the "FunnyBot" request, which involved using language models to generate jokes from predefined categories. The user's work included refining the problem definition, exploring dataset options and training setups, incorporating feedback, and clarifying the approach for data collection and model training. The focus appears to be on applying language models to humor generation and researching techniques to improve the model's ability to generate jokes.
deep-learningproblemsmojomachine-learning
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.