Yaroslav Kopotilov

This is a simple personal website powered by Flask and Dash. It is run in a Docker container and has monitoring systems tracking web activity and errors. While I'm not specialized in web development, the ability to create a simple web interface to visualize data or machine learning model predictions can be very handy.

This is a simple yet efficient Python package to parse XML. The package is written specifically for the fast extraction of tabular data (unlike xmltodict, which handles XML of any structure but slower). XML is not the most data science-friendly format, so the ability to transform it to Pandas or SQL can be very handy.

In 2018, before I started to work as a data scientist, I was studying textbooks on machine learning and testing the newly learned methods in various mini-projects. That's when I found this competition about predicting store sales on Kaggle. Time series is one of my favorite subjects, so I jumped in.
It was very fun to explore and visualize the dataset, to find interesting quirks in it. In particular, soon it became clear that this data had been synthetically generated, which gave out an important clue on how to solve this problem. And it was very exciting that in the end, my analysis paid off and I scored the first place!
Also, I was working on this project with my ex-colleague, so it was a good collaborative experience with just a touch of project management. Of course, it was far from the complexity of managing a real data science project—still, it gave me at least some sense of what might be waiting ahead.

An open-source Python package to create data pipelines based on the Dask package. It features lazy computation and cache loading, pickle and parquet serialization, and support for hashing of NumPy arrays and pandas DataFrames.

A Telegram bot powered by GPT with text-to-speech, speech-to-text, custom roles, and many other useful features. I contributed as a data scientist (GPT model benchmarking, prompt engineering, text embeddings), software developer (asynchronous Python code, the OpenAI API), and project manager (helping everyone to stay on the same page and deliver new features on time).

A collection of Jupyter notebooks exploring various Data Science topics. I created multiple examples of how to apply machine learning, deep learning, experiment design, efficient numerical computation, and visualization. This is a work in progress.