Working with Russian/English/Arabic texts to infer their sentiment and similarity (pytorch, transformers) and to cluster them into topics (sklearn, pyclustering). The NLP system allowed automating some business tasks. Development of models for sanity checking web crawling results (i.e.: broken encodings, CAPTCHAs, legal aggregator websites). The tech stack used: sklearn, pymorphy2, unicodedata, pandas, requests, ftfy, BeautifulSoup4, SQL, boto3/botocore. Interaction with product and web development teams. Maintaining CI/CD Pipeline on Github Actions. Working with SQL database. Measuring models' performance (creating guidelines for annotation, gathering unlabeled data, calculating precision/recall/accuracy).

Building and maintaining text preprocessing pipelines for English and Russian languages (includes: gathering/scrapping data, lemmatization, feature extraction, etc). Performing machine learning with jupyter+bash and sklearn/scipy/numpy/pandas. Developing and tuning topic models for English and Russian languages (BigARTM, gensim, topicnet). Developing topicnet library (open-sourced at GitHub, available at PyPi) for unsupervised NLP. Writing academic papers and technical reports (English and Russian), preparing papers for publication. Presenting at seminars and conferences. Creating a CD/CI pipeline (pytest, Bitbucket Pipelines/Travis CI), maintaining code coverage. Developing a backend for production machine learning solutions and demo stands (REST, Flask, Docker, MongoDB).

Developing and managing HFT trading bots, market data analysis, backtesting.

Development under Django (my team was developing a fork of edX platform for EdTech). Generation of problems and figures for school math education (using Python + matplotlib).

Java Development for Android (Pifagorea game, available as Android app on Google play).