Talks

FastAPI and Clean Architecture are a powerful duo for creating flexible and testable APIs, but it's important to properly organize application layers and manage dependencies through IoC.

Purpose of the talk: to show the application of evolutionary algorithms in engineering software. We will talk about how we at Sudo apply evolutionary algorithms in automatic design software. I will talk a bit about the problem itself, the classical approach to solving it, and how we improve it using machine learning.

We'll be summarizing the results, recall the highlights and talk about future plans. Come to the hall or join the broadcast so you don't miss a thing!

Flexible, fast and maintainable code for complex calculations is not magic, but the right approach. Let's see how to combine everything in Python on the example of petroleum physics.

How to use Taskiq to build microservice collaboration.

I will talk about our experience in developing a distributed map for unmanned transportation that provides a high level of accuracy, reliability, and scalability.

As the author of the FastStream framework, I'll walk you through the evolutionary development of the project's codebase, which you'll discuss during the talk. I propose you to build the framework design from scratch, make key architectural decisions, and deep dive into the code base of the project.

The master class will consist of two main parts: an introductory presentation and the competition itself. Each part will be accompanied by detailed explanations and an opportunity for participants to get actively involved in the process.

When the code base is 10,000,000 lines of code, you can't just put mypy --strict and be happy. You have to test everything, and you have to consider CI requirements and performance. So I took mypy, Pyright and red_knot and crisscrossed them with all versions of Python + JIT, nogil, tail-call and PyPy.

The talk is a complete guide to selecting a static type analyzer.

Why we decided to develop onETL — a library for ETL/ELT pipeline on top of Apache Spark. What we lacked in Spark and other tools, and how we fixed it. It will be useful for data engineers and those who develop tools for them to simplify their work.

I'll tell you about a hobby project: a night quest in the forest with work with light, sound and animators. It is implemented as a distributed service on ESP8266 + Wi-Fi + Bluetooth. Everything is written in Python: the embed, the control part, and the web interface — literally in a few days, most of it right in the forest.

An in-depth talk, based on practical production experience, on web scraping technologies and methodologies in the context of modern business challenges. We will cover the full development cycle of automated data collection solutions: from the planning stage to industrial operation. 

We will consider the transition from chaotic log collection to a full-fledged Observability stack that allows you to quickly find errors and react to them.

Breaking down the Airflow claims and dispelling the myths.

We used to say in junior interviews that Python's weakness is its inability to handle CPU-bound tasks. But it's easy to give an example where Python's performance is comparable to most compiled languages. Especially if we directly call modules written in them from our code.

How to simplify gRPC code generation for Python as much as possible. The main problem I faced: I needed a flexible tool to generate ready-to-use gRPC clients/servers with full typing support and using Ozon infrastructure.