Application. Source

Developed an algorithm for cryo-EM maps resolution estimation using deep learning.

• Collected and prepared training data, pairs of electron density maps, and local resolution maps.
• Developed and trained neural network model to estimate local resolution map based on electron density map using 3D‑UNet model architecture.
• Based on the trained model, created a CLI tool as well as a web application for cryo‑EM maps resolution estimation using deep learning.
• Skills: Python, Django, PyTorch, Keras, EMAN2, RELION-3, Scikit-learn, Bash.

Application. Source

Python Django-based personal portfolio website. All content: personal information, portfolio projects, social media links, etc. can be adjusted in Wagtail admin.

• Created a Django‑based personal portfolio website.
• The website uses Wagtail CMS. Wagtail is a Django Content Management System.
• Code from the attached repository can be used as a starting point for developing a production‑ready Django personal website with deployment to Heroku.
• Skills: Python, Docker, JavaScript, CSS, HTML, Django, Wagtail, SQL.

Results. Source

Content-based recommender system API based on the text of the post and user data.

• Created a content‑based recommendation system based on user data and post texts.
• Built a web API using FastAPI to predict the most relevant posts for a user.
• Added an A/B testing system to select the most appropriate recommendation model.
• Skills: Python, Docker, SQL, PyTorch, Scikit-learn, CatBoost, FastAPI, Optuna, NLP, Transformers.

Results. Source

The single-cell transcriptomics analysis has multiple steps, but we have focused on data integration — a crucial step when working with clinical data coming from patients.

• Researched libraries for scRNA‑seq data analysis and batch effect removal.
• Tested 8 algorithms on different data types and select the most suitable one for each case.
• Created software for batch effect removal from scRNA‑seq data using the best algorithm for the current case.
• Skills: Python, R, Scikit-learn, Scanpy, BBKNN, MNN, Scanorama, Cell Ranger, Bash.

Results. Source

This project aims to study differential genes expression of 19 sportsmen during physical and psychological stress before and after running in extreme highlands conditions.

• Processed and evaluated the quality of raw reads.
• Aligned processed reads on the human reference genome and counted gene and isoform expression levels.
• Identify the list of differential expressed genes and their functional analysis as well as cluster analysis.
• Skills: Python, R, DESeq2, FastQC, Bash, STAR, RSEM, MSigDB, GeneQuery.

• Python (Numpy, Pandas, Matplotlib, Seaborn, Sklearn, PyTorch, Keras, FastAPI, Django)
• R (ggplot2, Seurat, DeSeq2, dplyr)
• Linux, Bash, git, GitHub, Bitbucket, Docker, Kubernetes, Airflow, Jira
• JavaScript

• Classical Machine Learning (linear models, tree-based approaches, Catboost, LightGBM, XGBoost, Bayesian methods)
• Deep learning (MLP, CNN, image segmentation, detection, RNN, LSTM, Transformers, AE, VAE, GAN, TabNet)
• Model tuning (Optuna, genetic algorithm, Boruta)
• Interpretable machine learning (SHAP, LIME, Pixel Attribution)

• Hypothesis testing, ANOVA
• Survival analysis (Kaplan-Meier curves, Log rank test, Cox regression)
• Casual inference (Propensity score, Instrumental variables, Regression discontinuity)

• Databases (NCBI, UniProt, PDB, MsigDB, SILVA)
• Command-line tools (Cellranger, cellSNP, Picard, BLAST, GATK, STAR, SPAdes)
• Protein sequence analysis tools (MAFFT, MUSCLE, HMMER, ESM)
• Protein structure analysis tools (Rosetta, Phenix, Coot, AlphaFold)

• Russian – Native
• English – Full professional proficiency
• German – Elementary proficiency

• Biological education, which helps me to understand specialized biological and medical literature
• Work experience in molecular, microbiological, and biochemical labs

• Agile software development methods
• Presentation skills