Vladislav is available for hire

Vladislav Grozin

Verified Expert in Engineering

Recommendation Systems Developer

Location

Saint Petersburg, Russia

Toptal Member Since

March 1, 2018

Vladislav is a lead data scientist with an entrepreneurial mindset and a strong industrial background. Vladislav is especially interested in working on practical applications of data science and recommender systems or related areas like personalization, rankings, searches, and CTR predictions.

Machine Learning Recommendation Systems Python Python 3 Docker Spark R PostgreSQL Hadoop Elasticsearch PyCharm Jupyter PySpark Go Luigi

Portfolio

Constructor.io

Luigi, PySpark

Joom

Go, Scala, Spark, Python, PySpark

Diginetica

HDFS, PostgreSQL, Elasticsearch, Python

Experience

Recommendation Systems - 6 years Data Science - 6 years Machine Learning - 6 years Python - 6 years Spark - 2 years

Availability

Part-time

Preferred Environment

Jupyter, PuTTY, PyCharm

The most amazing...

...thing I've done was to research and develop a semantics-aware, typo-tolerant personalized text search engine for an eCommerce.

Work Experience

Senior ML Engineer

2019 - 2020

Constructor.io

Developed an IR system to handle search requests that were not handled by the main search engine by providing recommendations and query reformulations suggestions via a query expansion. PySpark and Luigi were used to write the pipelines.
Developed a recommender system component that extracts geography and time-based rules from user behavior and uses them for personalization. These rules improve offline ranking metrics and are interpretable; they were used in sales pitches.
Improved data-pipeline-testing procedures by implementing methods for mocking data pipeline inputs and outputs. This has led to better test coverage.

Technologies: Luigi, PySpark

Senior ML Engineer

2018 - 2019

Joom

Developed a system that optimizes interactions with users (discount coupons, push notifications) to increase user LTV. Thompson sampling was used to choose interaction parameters. Conversion was improved by 2%; Spark was used to process data.
Collaborated with universities on NER project. Worked out problem statement (NER of attributes in search queries), prepared data and worked with students on possible solutions. SpaCy-based model has improved coverage by 2X (over existing heuristics).
Developed a prediction model to automate marketing campaigns budget allocation. Increated the turnover threefold over the course of six months without dropping ROI.

Technologies: Go, Scala, Spark, Python, PySpark

Data Scientist

2016 - 2018

Diginetica

Served as the lead data scientist and product owner of a core company product, item search, and ranking for eCommerce. Elasticsearch was used for pre-fetch items and we trained a machine learning model to rerank fetched items.
Researched and developed a similar-and-supplementary item recommender system and a search product for eCommerce.
Analyzed A/B tests and tested various hypotheses for data-driven product development.
Wrote and published a paper on the development of a recommender system.

Technologies: HDFS, PostgreSQL, Elasticsearch, Python

Data Scientist

2014 - 2014

Data Mining Labs

Researched about prediction models for the ATM cash flow.
Wrote and published a paper about the ATM model.
Developed a web prototype to present the model to the customers.

Technologies: AngularJS, R

Experience

Recommender System Test Bench

https://github.com/rampeer/recommender_test_bench

Developed a test bench to try out different classic recommender systems.

Collaboration with VideoAmp

The goal of the project was to improve the forecasting of TV program viewership. It incorporates new content features which had been improved; R2 from 0.9 (current solution) to 0.93. A paper about the work done in this project was also written; it is going to be included in VideoAmp’s "2017 Recap" article.

Letter Personalization for an Email Delivery System

https://sendpulse.com/

This was a project for SendPulse.com, an email delivery system. The goal was to increase the open rate by personalizing letters by customizing the send time.

There were obstacles to consider:
• Sparsity: a majority of users have one or zero open letters
• Bias: the send time tended to be constant for each user which influenced and skewed dataset
• Various uncontrollable factors: the letter text/title also affected open rates

These obstacles were all avoided by our solutions:
• The development of a Bayesian subscriber behavior model
• User clustering for prior parameter selection
• Offline testing via a subscriber simulation

The production version was launched as a microservice by using Python REST (Flask and Unicorn) and MySQL in Docker containers; it has shown a 5% relative increase in the open rate versus the existing AI solution.

Education

2015 - 2017

Master's Degree in Computer Science (Corporate Information Systems)

ITMO University - Saint Petersburg, Russia

2011 - 2015

Bachelor's Degree in Computer Science

ITMO University - Saint Petersburg, Russia

Skills

Libraries/APIs

PySpark, Luigi, Flask-RESTful

Tools

Docker Compose, PyCharm, PuTTY, Jupyter

Paradigms

Data Science

Languages

Python, Python 3, Go, Scala, R

Platforms

Docker

Storage

HDFS, Cassandra, PostgreSQL, Elasticsearch

Frameworks

AngularJS, Spark, Hadoop

Other

Machine Learning, Recommendation Systems, Business Analysis

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring