Student of Financial University under the Government of Russian Federation
Faculty of Information Technologies and Big Data Analysis (ITiABD).
Degree: Applied Information Systems in Economics and Finance.
Average score - 94.808
Rating - 8/221 (after 3 semester)
Python, SQL, Machine Learning, Git, Airflow, Mlflow, Data Science, Linux, Mathematical Analysis, Pandas, Numpy, Math Statistics, BI Analytics, Teamwork, PostgreSQL, PyTorch, TensorFlow, Docker, Kubernetes, CI/CD, PySpark, Hadoop
- Field - Applied MLOps
- Worked with PySpark, Docker, Kubernetes, CI/CD processes, Airflow, MLflow
- For the final project, a cookiecutter template was created for data scientists to compare a retrained model in production with an AutoML script (PyCaret) https://git.angara.cloud/a2p-demo-prod/a2p-demo-prod/project-trainee/automl-dag
- Served as the project architect
- Conducted dataset analysis and identified key insights
- Designed a solution to increase SKU profitability for P&G by transferring information from merchandisers' phones (OpenCV) to Airflow for data orchestration, followed by the construction of an optimal store planogram
High quality award - 25%
- Developed initiatives for the introduction of children's products into the company to increase the company's margins
- Tested hypotheses and highlighted the most promising initiatives in the short and long term
- Built a financial model for possible case solutions
- Development of a model for determining the location of openings for exterior envelopes (cells)
of a building under construction on the basis of photos of its facade, and determination of the degree of readiness of
the corresponding
cells (yolo v8 model) - Subsequent implementation of docker container to run the program on a remote server.
Implementing a machine learning algorithm to predict the price of goods for the next 90 days for 5 cities, subject to the following conditions
- One price for a product must be held for more than 3 days (i.e., it is impossible to set a price for one day and change it the next day).
- it is forbidden to change the price by more than 1 gold at a time (i.e. you cannot change the price from 3 gold pieces to 4.50 gold pieces, but you can change the price from 3 gold pieces to 4.50 gold pieces and then 3 days later to 4.50 gold pieces. gold and then raise it to 4.5 gold 3 days later)
- Don't be tempted to set the price too high - everyone will refuse to buy from you, And if the Ancient Gods notice that your price is 20% higher than your competitors', they can may punish you for your greed (with a heavy fine).
has implemented such machine learning models as GRU, SARIMA, Random Forest, Gradient Boosting, SVR, and LSTM as a final
choice with the best accuracy.
(You can see the report and the models themselves in the Gazprom folder)
- Developed models for forecasting the state of power grid facilities and electricity consumption by organizations and enterprises (lstm, gradient boosting, snn, spiking nn) with further analysis of the impact of forecast values on technical and industrial facilities.
- Developed models for forecasting the water level of the Republic of Bashkortostan (gradient boosting, polly regression, RNN, SNN) with site realization on django.
- Data mining to analyze real estate market needs
- Parsing of the company's website using Python(BS4, json, xml, requests)
- Autonomous loading of objects into CRM system
- Creating presentations (Figma) for clients
Graduated with honors from the grant track of the School of Analytics
Graduated with the 'B' diploma from the case school