Data scientist with strong programming and academic background, interested in data mining using machine learning and other AI techniques.
Read More »
I've studied machine learning and data analysis online courses from leading universities (Stanford, Duke, Caltech) and trained on practical projects from Kaggle with a score in the top 10% of involved data scientists. I've recommended missing links in a social network (Facebook), predicted the salary of job ad based on its contents (Adzuna), auction price (FastIron), built a web pages classifier...
Sep 9, 2015|Data Analysis|Private|Completed
Jun 13, 2015|Data Science|Private|Completed
May 26, 2015|Data Analysis|Private|Completed
My projects concerning machine learning and data mining:
Machine learning for advertisement targeting (Video International Group). Predicted gender and age from the user's click stream. Investigated the possibility to predict the type of TV viewer from the user's click stream. Used tokenization, latent semantic analysis (LSA), logistic regression or random forest.
Tools: python, sklearn, scipy, numpy, nltk, Cassandra, Hadoop
Recommended missing links in a social network (Facebook) as a Kaggle competition. Earned 23rd position out of 418 data scientists (top 10%). The objective was to predict new links to users of large network (1.5M users, 10M links). Used some generative algorithms to predict nodes ratings: user-user recommendations, item-item recommendations, variations of PageRank. Combined these ratings as features with more features, mostly distances. Used random forest for learning.
Concepts: Collaborative Filtering, PageRank, Random Forest Classifier
Tools: python, sklearn, numpy, pypy, numba
Predicted the salary of UK job ad based on its contents (Adzuna) as a Kaggle competition. Earned 21st position out of 289 data scientists (top 10%). The objective was to predict salary from the contents of the job advertisement. Used natural language preprocessing methods (tokenization, stemming, lemmatization, TF-IDF,..). Different methods of supervised learning were used and blended in final prediction.
Concepts: Feature Selection, RF, Gradient Boosting, Blending...
Read More »
Freelance at Kaggle, Experfy, CrowdANALYTIX, etc.
2012 - Present
- machine learning for various projects:
- recommended missing links in a social network (Facebook)
- predicted the salary of UK job ad based on its contents (Adzuna)
- predicted the auction sale...
Senior Software Engineer
1999 - 2001
- startup: advertising system from scratch (ads management, ads rotation, realtime statistics, wed interface)
- core of this advertising system works at Rambler so far