- Train NLP models for any tasks, such as classification, text generation, summarization, topic modeling
- Train text2image models, such as Dalle-2, Stable Diffusion.
- Integrate any models in services and pipeline.
- Train classical models, such as boosting (xgboost, lightgbm, catboost)
- Build ranking models and search
- Collect datasets from web.
- OZON.ruTeam LeadDIGITAL & ITAugust 2022 - Today (2 years and 4 months)Moscow, Russia1. Develop microservice architecture for real-time matching2. Implement algorithms for finding products which are not represented on the marketplace3. Optimize production pipelines4. Leading a team of 6 ML engineers5. Communications with customers
- OZON.ruMachine Learning EngineerDIGITAL & ITSeptember 2021 - August 2022 (11 months)Moscow, RussiaDevelop ML Matching algorithms, which find the same products from a competitor marketplace.1. Implement several ML model for products matching with precision 95% and recall 25%2. Train and integrate in pipeline several NLP models (BERT, fasttext, BM25)3. Implement a lot of algorithms analytics in Grafana and Slack.pySpark, Hadoop, Kafka, Airflow, Grafana, PyTorch, NLP
- SberbankData ScientistApril 2021 - September 2021 (5 months)St Petersburg, RussiaImplement Recommendation System based on client-manager dialogs. Spark, Hadoop, Pytorch, NLP
- Master of ScienceУниверситет ИТМО2022Master's degree, Data Science
- Bachelor's degree, Математика и компьютерные наукиУниверситет ИТМО2020Bachelor's degree, Математика и компьютерные науки