Our team develops the stream processing platform for Yandex's advertising network. It is based on BigRT and YTsaurus. We deliver data with sub-minute latency and exactly-once guarantees, processing 2 million events or 5 GB per second with trusted accuracy. This is necessary to withstand constantly growing loads, ensure fault tolerance, and improve data processing quality and speed.

We face technical challenges: * Improve the data access API * Make adding new data simple and convenient * Prove that ClickHouse can be replaced with YTsaurus dynamic tables for our tasks * Become more reliable and improve performance

What tasks await you

Development of GRPC API Replacing the current HTTP API, simplifying and expanding it.

Support for asynchronous API We are currently considering implementing a task queue using technologies close to us. As a result, we will support an API for generating reports via a queue.

Simplifying the addition of new data Adding a new log, field, or column should be simple, predictable, and not lead to breakage. The tools should be convenient.

Ensuring reliability and performance The system must transparently survive data center outages, manage to process accumulated backlogs in case of failures—and all this despite architectural complexity, feature rollouts, and new data.

We expect you to

• Want to write code in C++
• Know classical algorithms and data structures
• Be familiar with Linux

It will be a plus if you

• Are ready to lead a small group of developers
• Have written code using the asynchronous paradigm
• Can program in Python
• Have developed network applications
• Have experience in stream data processing