Schedule

The quality of medical AI cannot be measured by a single offline metric — the error may lie in the data, configuration, preprocessing, integration, or infrastructure of the client. Using real files, I'll show you how we test the entire research path and why we consider monitoring to be part of testing.

How to test autonomous multi-agent systems if their quality is not deterministic and errors can hide deep within the chain of agents?

Real experience with Java and Spring: OpenCode and Qwen 397B in a closed loop, console in the frame, not a single manual edit in the case shown.

Taking the same UI scenarios, we will compare different approaches to AI‑based test generation using Playwright CLI, Journeys, Maestro, and other tools, and we will also calculate the cost of creating, running, and maintaining tests.

At the workshop, we will build our own QAI agent, an agent system that helps us test tasks and perform root-cause analysis.

Agents in QA work: from bug analysis to quality gate in pipeline. What is hidden behind each setting, how much does one dialogue move cost, and why do the Playwright authors advise against using their own MCP.

Agents break code when refactoring not because the model is weak but because it was handed a text editor instead of tools with guarantees.

We'll look at what happens when you give an agent an IDE's refactoring engine, how to design tools like these, and where the approach has its honest limits.

At the workshop, we will discuss how modern agent systems are structured and what techniques are used to test them for security.

Using the example of a real multi-agent autotesting system, we show how choosing the right model for a task and several engineering techniques result in 40-90% savings without loss of quality.

We will analyze real cases with our experts, debate the new role of a tester, and separate working practices from futuristic hypotheses.

How to find bugs not in applications, but in the QA tools themselves, which we trust to provide testing results.

The quality of medical AI cannot be measured by a single offline metric — the error may lie in the data, configuration, preprocessing, integration, or infrastructure of the client. Using real files, I'll show you how we test the entire research path and why we consider monitoring to be part of testing.

We analyze real‑life cases from practice without prepared answers: on stage, experts will take turns proposing their solutions, explaining their thought process, and justifying their approaches.

Let’s talk about how to assess real‑world conditions, identify dependencies and limitations, and make decisions that help the team create a system that meets the market needs.

What does product quality look like in production, and what can QA do to ensure its reliability? Let’s talk about the SRE approach in testing: SLI/SLO/SLA, Observability, Error Budget, Toil, fault tolerance, and incident management.

We’ll look at practical examples to see how QA can influence not only the quality of the release but also the stability and reliability of the product.

Using the example of a real multi-agent autotesting system, we show how choosing the right model for a task and several engineering techniques result in 40-90% savings without loss of quality.

A real story of how my team went from 30-40 minutes per integration test run down to 3-4 minutes. We'll talk about how we got there, the mistakes we made along the way, and where we ended up.

We analyze real‑life cases from practice without prepared answers: on stage, experts will take turns proposing their solutions, explaining their thought process, and justifying their approaches.

We will analyze real cases with our experts, debate the new role of a tester, and separate working practices from futuristic hypotheses.

What should be the architecture of an autotest framework for the backend? Let's discuss proven solutions, scalability, and practical experience.

How to automate the analysis of automated test failures after CI/CD runs without turning this process into “we’ve handed everything over to neural networks”?

We will be happy to share our experience in developing performance testing tools, interacting with vendors, and selecting basic technologies.

At the workshop, we will build our own QAI agent, an agent system that helps us test tasks and perform root-cause analysis.

As the pywinauto maintainer, I talk about cross-platform desktop UI tests, about code (DLL/SO) injection into UI process which should challenge commercial tools, and explain few complicated cases.

In this talk, we’ll explore several everyday scenarios. We’ll prepare a device; change its language, theme, and network settings with a single click; install an APK; and identify the exact build version.

Agents break code when refactoring not because the model is weak but because it was handed a text editor instead of tools with guarantees.

We'll look at what happens when you give an agent an IDE's refactoring engine, how to design tools like these, and where the approach has its honest limits.

I will share a practical case study of Data Quality implementation in a tax reporting project, where the synergy of PyDeequ and Soda Core helped stabilize calculations and prevent financial risks. Additionally, I will showcase our custom DQ-generator, which transforms metadata into ready-to-use validation rules and significantly reduces engineers' routine work.

I’ll tell you how and why we at Ozon created our own Testo framework for testing in Go and released it as open source.

How to automate the analysis of automated test failures after CI/CD runs without turning this process into “we’ve handed everything over to neural networks”?

Using practical examples, we will discuss how to review the test model, find blind spots and focus on checks that really reduce the risk of defects in production.

Using small, visual examples, we will explore typical "coverage illusions" (green strings and increasing percentages that do not improve the quality of tests), show how cyclomatic complexity can set a lower bound on the number of tests required, and discuss mutation testing as an additional approach and its practical limitations.

We will tell you how we run end-to-end tests on production code at Avito without a full E2E environment, isolating external dependencies using fakes.

What should be the architecture of an autotest framework for the backend? Let's discuss proven solutions, scalability, and practical experience.

I’ll tell you how and why we at Ozon created our own Testo framework for testing in Go and released it as open source.

What is field testing of telecom equipment and why is it needed in a real network?

How to flexibly and painlessly automate CLI testing using FW as an example.

In this talk, we’ll explore several everyday scenarios. We’ll prepare a device; change its language, theme, and network settings with a single click; install an APK; and identify the exact build version.

At the workshop, we will discuss how modern agent systems are structured and what techniques are used to test them for security.

We will be happy to share our experience in developing performance testing tools, interacting with vendors, and selecting basic technologies.

A networking session before the start of the conference day. Here you can easily meet other participants, find your first professional contacts, and start the conference with some conversation.

The results of the TechRadar survey.

TechRadar shows what is happening in the industry. And we offer to understand together — why exactly so and what it means for our work. After the presentation of the study, we invite you to discuss the most interesting results at the round tables together with experts and colleagues. We will collect and present the main conclusions and insights of the participants at the closing of the conference.

Let's talk honestly about testing failures: real stories, useful insights, and no recording. Do you have your own epic failure? Fill out the form and tell us about it!

A networking session before the start of the second day of the conference. An opportunity to meet new participants, continue conversations that began the day before, and find new professional contacts.

Let's trace how the approaches to testing quality assessment have evolved: from line coverage and data coverage to fuzzing, context coverage, and business coverage.

Bring to the discussion what’s happening at your workplace right now: a difficult situation in the team, a block in the process, a conflict, an unclear bug, or a question that hasn’t been answered yet.

Let's discuss the results of modern psychological research and find out why critical thinking is becoming especially important in the age of AI.