The ability to add users as friends is one of the most important mechanics of any social network. The vast majority of interactions occur between users who are friends with each other. They see and comment on each other’s posts in their news feeds and go to their friends lists to start chats. This is why the growth of the social graph is so important.

My name is Evgeny Zamyatin. I’m part of the Core ML team at VK. …


Hi, my name is Daniel. I’m a member of the Core ML Team at VK.com. I work on recommender systems, which is also the topic of this post.

An interesting paper was presented at NeurIPS 2020: https://papers.nips.cc/paper/2020/hash/019fa4fdf1c04cf73ba25aa2223769cd-Abstract.html.

The main idea of it is that if you somehow preprocess the training data for your matrix-factorization based recommender model, it can yield significant metric improvements on the test set. For example, here are the most impressive results from the paper’s supplementary material:


Hi, everyone. I’m Anver from the CoreML team at VK, where I design and improve algorithms for news feed ranking.

Improving ranking algorithms usually results in improving the click-through rates(CTRs) of users’ actions (e.g., CTR of likes, CTR of comments, etc.). We will refer to these metrics as short-term metrics.

But there is another set of metrics, which include time spent and number of sessions per user. These metrics are considered to be far more significant, since they’re believed to be much better measures of users’ satisfaction with VK.We will refer to these metrics as long-term metrics.

Optimizing long-term metrics…


Abstract

Hello, we are researchers from ML lab at ITMO University and CoreML team at VK. Automatic post classification is an important task for VK. It’s used not only to compose thematic feeds for users, but to also identify inappropriate content. Assessors are involved in solving this task. The cost of using them can be significantly reduced by using machine learning techniques such as active learning. In this article, we will discuss the use of active learning in classifying multimodal data. …


Hi, we are Nikita and Daniel from the CoreML team at VK. It’s our job to design and improve recommender systems for friends, music, videos and the news feed. This involves tons of A/B tests, and our progress at the end of the day relies heavily on how accurate and efficient our A/B tests were.

The two most essential things in A/B tests are the design of the experiments and accurate analysis of the experiments’ results. In this article, we will stick to the most common design and compare various statistical analysis procedures, from the very standard t-test and Mann-Whitney…


Libraries for design tools are an important part of a design system. We believe that designers should work with elements that are as close as possible to what is already available in code. That is why we’ve created libraries containing all of ourmost up-to-date components, which reflect the states available to developers.

Our developers have already published the source code for React components on Github, so we decided to follow their example and share our Figma component libraries.

There are four libraries available on the VK Figma Community Page:

  • VKUI Base Library, containing over 650 icons and all color tokens…


VK has one of the largest PHP codebases in Russia. When you have several million lines of code, even seemingly simple refactoring tasks can turn into an exciting journey. Put a large monolith in a monorepo, add several tightly connected system parts, and even the smallest change couldrequire your full self-confidence to merge.

NoVerify is a static analysis tool that helps VK developers. Not only does it assist in finding potential problems in the code but it also helps maintain a consistent programming style. This past spring, the linter was made available in opensource on GitHub.

Of course, we tried…


We want to introduce our new text tokenization tool: YouTokenToMe. It works 7 to 10 times faster than other popular tools for alphabetic languages and 40 to 50 times faster for logographic languages. Here we’ll tell you about YouTokenToMe and share this open-source tool with you on GitHub. You can find a link to it at the end of the article.

Today, the majority of tasks handled by neural networks have to do with text processing. However, neural networks work with numbers, so the text needs to be pre-processed before it can be fed to the model.

The following methods…


How the photo editor for Android transformed from Snapster’s initial version to official VK app filters.

This story begins in 2015 when we launched development efforts for Snapster, an independent app for editing photos. At that time and in our official apps, as well as in Instagram, photo editing tools were fairly primitive. Take a picture, make simple color modifications using OpenGL-shaders and apply textures on top if you want glamourous images. Challenge accomplished.

After several months passed since development began, when the first version of Snapster was ready, it still used filters taken from the main VK app. …


Everyone knows that the heart of VK is the Singer House on Nevsky Prospect. Now you will learn where the brain is located and how it works — that is, the ICVA data center.

How is the data center built?

The data center, a facility for storage and data processing, is a collection of several infrastructure systems that provide reliability and fault tolerance for server work and network equipment.

It is impossible to just install a bunch of servers and switches. But rather one must create and maintain optimal conditions for work. …

VK Tech

About vk.com technologies and infrastructure

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store