Download PDF

Bartek Roszak

Data Scientist at Inbox Liberator

Summary

Passionate about new technologies, especially about AI field.

Inborn analyst.

Quick-learner.

Driven in the pursuit of my goals.

Work experience

10-2017Present

Data Scientist / NLP Engineer

Inbox Liberator

Designing whole learning system to create state-of-art text classification model.

- using machine learning techniques to choose appropriate data for training to achieve the best generalization
- managing a team responsible for providing high quality data for training
- building, mixing and stacking various neural net architectures (CNN, LSTM, GRU etc)
- creating and deploying deep learning models to check if data is correctly labeled
- creating algorithms to choose models which will be the best to stack

Main tools:
Python: PyTorch, sklearn, nltk
R: t-sne, h2o, tm, 
openNLP, coreNLP, word2vc, glove, fasttext

08-2012Present

Trader

Coco Trading

Creating Machine Learning models for auto-trading.

Company's capital managing (1 000 000 USD).
Creating simple decision trees for automating trading.
Searching for correlation between stock market and derivatives.
Co-creating analytics tools.

11-201610-2017

Data Scientist

Transparent Data

Creating machine learning models.
Implementing these models on production.
Recommending new technologies which can replace present solutions.

Main projects:
Image processing (R + h2o + tensorflow)
- character recognition
NLP (Stanford's coreNLP, Apache's openNLP, R)
- entity idetification (finding company name in press article)
- text classification (unification of companies representation types)
Unsupervised learning (R + RMarkdown)
- customer segmentation
Record Linkage (Python + Record Linkage + ElasticSearch)

Education

09-2016Present

Advanced analytics with SAS Institute

Poznań University of Economics
10-200606-2010

Finance and Accounting

WSB Universities of Economics