Testing a financial trading platform means combining strict precision requirements, real-time data, shared environments, and parallel execution at scale. In this session, Aidos Kenessov will walk attendees through how the QA team at EXANTE - a European multi-asset broker - built and evolved a Python and pytest test automation framework covering dozens of back-end microservices.

The talk is structured around four practical pillars: layered test architecture with strict naming conventions, idempotent test data through the Setupers Pattern, reliable execution via smart retry decorators and parallel pytest-xdist runs, and end-to-end observability connecting Slack alerts to Allure reports and Jaeger distributed traces.

Attendees will see real code examples from the framework, a war story of how blanket test reruns nearly destroyed team trust in regression results, and concrete metrics - including how a combination of file splitting, retry decorators, and smart waits brought a critical test suite from 20–25 minutes down to two minutes, making it 8.5 times faster.

By the end of the session, attendees will understand how to organize a large test suite using clear architectural layers, how to design test data that survives shared environments and parallel execution, how to replace flaky `sleep()` calls with deterministic waits, and how to instrument tests with OpenTelemetry so any failure can be traced across microservices in seconds. The session emphasizes patterns that scale - applicable beyond fintech, in any domain with multiple services and strict reliability requirements.