Dmitriy Kravtsov
Purpose: get the position of Data Scientist, ML Developer, ML Engineer
Place of residence: Odessa, Ukraine
Phone: -
E- mail :-Desired salary: from USD 2500 per month
Skills:
●
●
●
●
●
Tabular Data: python, numpy, matplotlib, seaborn, pandas, sklearn, SQL
NLP: nltk, BERT, TF-IDF, GloVe, text summarization and classification
Time Series: interpolation, autoregression, FB Prophet, VAR, SARIMA
Computer vision: Tensorflow, Keras, CNN
English: strong intermediate
Basic and additional education:
●
●
●
●
Machine Learning and Data Analysis, Coursera, Yandex, MIPT-
Udemy: Python 3: complete guide 2022
Kaggle Data Science Courses,- (Deep Learning, Feature Engineering, Pandas, Python,
Machine Learning, NLP, TimeSeries, Image classification)
Odessa National Maritime University, faculty of "Transport technologies and systems"; specialty "System analysis and Logistics", master
Experience:
ML Engineer, Developer
August 2021 - present time
●
●
●
●
●
Development and update Catboost, XGBoost and LGBM based models for buy/rent cost of house
price prediction in the United States with a MdAPE around 5%
Statistical and dynamic feature engineering, K-means and DBSCAN customer clusterization
Time Series based forecasting the dynamics of real estate prices with macroeconomic factors
(Linear and Polynomial Regressions, VectorAutoregression, SARIMA, FB Prophet with exogenous
factors, interpolation, savgol filter)
NLP based classification of Customers (classification by description with vectorization (TF-IDF,
GloVe) and modeling (MultinomialNB)), Property Address mapping, text summarization and
classification
Rooms images classification (Transfer Learning) through more than 20 classes with F-score 91%
Self-employed/Junior Data Scientist
April 2020 - July 2021
●
●
●
●
●
●
Applications based on such technologies like python, sql, flask, docker, docker, html, css, sklearn,
pandas, numpy and others have been developed by me.
Apps purpose: can be useful for real estate organizations involved in the sale of real estates
This applications allows users to identify real estate objects that are undervalued or overvalued
by their price(demo - http://odessapricepredictor.herokuapp.com/ )
As predictive models were used RandomForestRegressor, LinearRegression and XGBoost.
A description of the data processing and machine learning models is available at
https://github.com/dmkravtsov/odessapricepredictor/blob/master/api/project.ipynb
The system code of the project is available https://github.com/dmkravtsov/odessapricepredictor
Hobbies: Tourism, bicycle, reading, healthy lifestyle, fishing, tennis