Kan vara tillgänglig
(Uppdaterat 2023-09-19)Senior Data Scientist / ML Engineer
Helsinki-Uusimaa, Фінляндія
Modersmål English, Ukrainian, Nybörjare Finnish
- NLP, CV
- Python
- PyTorch
Kompetenser (28)
Machine learnign
K-MEANS
TENSORFLOW
Scikit-learn
COMPUTER VISION
NLP
Credit Risk
STATISTICS
Git
Python
OPENCV
SQL
Agile
Pytorch
Jenkins
Deep Learning
Linear Discriminant Analysis
AI
AWS
IMAGE PROCESSING
Kanban
Transformers
Machine Learning
LIDAR DATA
SIGNAL PROCESSING
Google Cloud
MATLAB
C/C++
Sammanfattning
I’m a machine learning engineer with more than 6 years of experience and a solid math
background, specializing in NLP, Computer Vision, time series, and recommendation
systems. Passionate about creating new and innovative solutions.
Professionell bakgrund
2021-08 - Pågående
● Developed object detection and classification models of private objects (eg credit cards) using TensorFlow, EfficientNetV2, YOLO with an F1 score of 0.96
● Object positioning, pose estimation and object tracking, camera calibration, and image processing (e.g., background subtraction, photogrammetry, image distortion removal)
● LiDAR point cloud data processing
● Developed a GPT-2 based model using PyTorch for detecting code vulnerabilities and fixing them automatically, later optimizing it with DeepSpeed, Quantization to reduce training model time by 50%
2020-07 - 2021-08
● Developed text classification BERT model combined with text clustering SBERT + LDA for news articles. Developed REST API to connect with MongoDB and MSSQL
● Implemented automating document parsing system for invoice data recognition with OCR, PyTorch, LSTM, FastText
2019-07 - 2020-07
● Designed and developed forecasting models for Fortune 100 companies in CPG and SRM areas using XGBoost, LightGBM, CatBoost, Faiss, LDA, TF-IDF
● Developed time series forecasting with Pyramid Arima, Prophet, Holt-Winters, TBATS, Hybrid CNN-LSTM
2018-02 - 2019-07
● Developed application, behavioral, and transaction scorecards models with scikit-learn, TensorFlow
● Created a strategy to manage the credit risk parameters (e.g. time series of delinquency payments, macroeconomic, default prediction modeling) with seasonal_decompose, Arima
● Created stored procs, functions, views, indexes in SQL to create a dataset for modeling
2017-07 - 2017-12
● Created time series models for sales forecasts using statistics, Python
● Automated and optimized processes of building reports reducing time of building by 30% with Python, PL/SQL
Akademisk bakgrund
2017-09 - 2019-07
2013-09 - 2017-07
Certifikat
2020-02
2021-11
2022-06
2020-03