Kotlin Analytics Team

I lead the analytics infrastructure team. We maintain one of the largest Clickhouse clusters in the country (96 servers) and also develop services for data analysts. This list includes (but not limited to) the following services:- MapReduce-like API for Clickhouse which helps data analysts to handle large data sets (billions of rows, TBs scale). This API is easy-to-use and doesn't require special knowledge of big data handling patterns. The service is used by more than 60 people every week (see details here https://youtu.be/fEHcHTZ378M).- A/B testing platform including front-end on RShiny. Our platform is highly scalable, we can summarize results of 200-300 simultaneous experiments with hundreds metrics in each. Instead of standard statistical tests, we use linear regressions and diagnostics based on the conditions of the Gauss-Markov theorem.- Anomaly tracking system based on Twitter's SH-ESD. We have refined this algorithm for processing streaming data.My role in these projects is primarily in system design, but I still contributed to our repositories as a developer. My activities not limited to technical ones:- I actively participate in hiring data analysts by conducting technical sections (more than 50 sections each year)- I organize internal seminars and meetups on data processing tools and AB testing. Our goal is to share best practices across business units of company.- as a product analyst, I participate in significant product changes. For example: a) redesign vk.com (web-version), DAU 5 mln, MAU 50 mln b) switching from http 2 to http 3 (QUIC) - it was series of 8 A/B experiments. It affects 60 mln on daily basis and more than 100 mln monthly c) redesign of user's profile on mobile platforms (iOS, Android) Project stack: Clickhouse, Postgres, Redis, RQ, fastAPI, numpy, pandas, statsmodels, RShiny

In this role, I was engaged in the development and design of a number of analytical services, for example:- the first versions of the tool for summarizing the results of A\B tests, including the original algorithm for preparing data for vectorized calculations (see section "Vectorized Welch t-test" here https://is.gd/cuTAOf )- anomaly tracking service based on its own SH-ESD patch (https://is.gd/oJoDxt )

Quantor.me is an ambitious startup consisting of 5 people, including me. The initial goal was to provide customers with pre-aggregated blockchain data through a useful API. Finally, It was a success story, because startup was acquired by 1inchI was involved into the following:- co-designing and co-implementing scalable blockchain (Ethereum) parsing software - setting up streaming infrastructure for DWH (Kafka + ClickHouse)- designing and implementing data-transformation utility Project stack: RQ, Kafka, Clickhouse, Nebula-Graph

I worked in business unit "Education". In this position, I focused on the following tasks:- setting up ETL processes on the internal infrastructure of the company- formulation of the key metrics of the product- setting up reporting, including dashboards with a complex backend on RShiny- recruiting both for the project and for other business units of the company

As a consultant, I participated in a company project with a large U.S. investment fund. My task was to provide consulting support on the R language, in particular RShiny, OOP (R6)

In this company, I performed a wide range of tasks from reporting to machine learning tasks (classification, time series). The most significant achievements:- the introduction of Clickhouse as an analytical DBMS- the introduction of Luidgi as an ETL tool- the company's revenue model is based on VAR/VECM, which made it possible to explain the negative phenomena when purchasing traffic

In this role, I performed reporting tasks using R, BigQuery, Tableau and Salesforce API