Hands-on notes on API and mobile test automation with Kotlin, Spring Boot, and Python—plus practical uses of AI in test design, debugging, CI, and regression.
API automation with Kotlin, Spring Boot, and Python.
02iOS and Android testing—from traffic analysis to UI regression.
03AI-assisted test design, failure analysis, and CI workflows.
February 2025 — August 2026
API, mobile, AI-assisted testing, and test architecture.
What worked in practice: Claude-assisted scenario design, synthetic state preparation, and evidence review for stable regression flows—while end-to-end checks stayed separate.
How I combined Claude, plain-language scenarios, synthetic fixtures, and CI quality gates into a reusable iOS testing workflow.
How I connected AI-supported test design and triage to selective GitLab runs, Allure traceability, scheduled coverage, and human release decisions.
How I used a curated QA knowledge base to design a modular Kotlin/Spring Boot test platform, with AI extracting reusable contracts and humans owning architecture, security, and business assertions.
A risk-based decision guide for Espresso, Compose testing, UI Automator, Kaspresso, Appium, and Maestro across native, system, and cross-platform flows.
How Proxyman streamlines iOS traffic inspection, response stubbing, and network-failure testing—and when Charles remains the better team choice.
How to organize XCTest and XCUIAutomation around stable accessibility contracts, screen objects, deterministic state, and clear visual-testing boundaries.
Schema validation and consumer-driven contracts overlap, but neither is a substitute for the other—or for a small number of real integration tests.
A practical way to model changing dependency behavior, inspect outgoing requests, and keep service-level tests deterministic.
Concurrent checks can expose coordination defects and reduce wall time, but only when client reuse, pool limits, timeouts, and failure handling are explicit.
Fast database tests are useful only when commits, identity maps, and shared state cannot make one scenario influence the next.
Pydantic can turn response validation into readable Python, but only when coercion, extra fields, and business rules are deliberate choices.
An API suite becomes more valuable when it verifies protocol meaning, representation contracts, and failure behavior—not just that every route returns something.
A strong parameter matrix exposes risk and boundaries. A weak one creates dozens of nearly identical failures with no explanation of what each row protects.
A fixture is most valuable when it makes ownership, lifetime, and cleanup explicit—and proves that every test starts from a state it can trust.