QA Automation Engineer Interview Questions (1970)

Designs, writes, and maintains automated tests (UI + API + DB) that run in CI to catch regressions before release. Unlike an SDET, the focus is heavier on test authoring and pipeline stability than on framework/infra R&D.

• SDET: builds framework, tooling, test infrastructure; codes like a dev
• QA Automation Engineer: consumes the framework, owns coverage for features
• Salary: SDET typically 20–40% higher; both are automation-first roles

Selenium 4 / Playwright, TestNG or JUnit 5, REST Assured or Postman/Newman, Git, Jenkins/GitHub Actions, Allure, Docker basics, SQL, and at least one cloud grid (BrowserStack, LambdaTest, Sauce).

Follow the pyramid: cover happy-path E2E in UI, cover edge cases in API, cover logic in unit tests owned by devs. Automate stable, high-value regressions first; leave exploratory paths manual.

• Tests that change every sprint (unstable UI)
• One-off exploratory scenarios
• Visual/UX judgment calls
• Anything where authoring cost > 10× manual time

public class LoginPage extends BasePage {
  private final By email = By.id("email");
  private final By pwd   = By.id("password");
  private final By submit= By.cssSelector("button[type=submit]");

  public HomePage loginAs(String u, String p){
    type(email, u); type(pwd, p); click(submit);
    return new HomePage(driver);
  }
}

Page Factory uses @FindBy annotations and lazy proxy init. In 2026 most teams have moved to plain POM because Page Factory doesn't play well with Selenium 4's new locator APIs and has hidden StaleElement issues.

Priority: data-testid > ID > ARIA role+name > unique CSS > relative XPath. Never index-based XPath like (//div)[7] — it breaks on the first UI tweak.

findElement returns one WebElement and throws NoSuchElementException. findElements returns a List, empty if none. Use the second for optional/conditional checks so you don't wrap everything in try/catch.

Wait for options to load, then use Select for native, or click-and-filter for custom. For React combo-boxes, use role selector: page.getByRole('option', { name: 'India' }).click().

Iframe: driver.switchTo().frame(...), then defaultContent() to exit. Shadow DOM: use JS executor arguments[0].shadowRoot.querySelector(...) in Selenium; Playwright pierces shadow automatically.

Implicit — global, applies to every findElement (avoid; hides real issues). Explicit — WebDriverWait on a condition, use this. Fluent — explicit + polling interval + ignored exceptions. Never mix implicit and explicit.

new WebDriverWait(driver, Duration.ofSeconds(10))
  .until(ExpectedConditions.visibilityOfElementLocated(loc));

Replace every sleep with a condition wait (visibility, invisibility, url-contains, custom lambda).

• Race conditions → wait on state, not time
• Shared test data → isolate per test
• Animations → disable via CSS override in test setup
• Third-party ads/analytics → block via network intercept

TestNG: IRetryAnalyzer with max 2 retries. Playwright: retries: 2 in config. Never retry more than 2× — beyond that you're masking a bug.

TestNG: parallel="methods" thread-count="4" in testng.xml. Playwright: workers: 4. Ensure each test creates its own data and uses its own driver/browser context.

Local: matrix of Chrome/Firefox/WebKit in Playwright config. CI: BrowserStack/LambdaTest via their REST hub URL. Only run cross-browser for critical smoke, not full regression — cost multiplies.

given().contentType(JSON).body(user)
.when().post("/api/users")
.then().statusCode(201)
.body("id", notNullValue())
.body("email", equalTo(user.email));

REST Assured: matchesJsonSchemaInClasspath("user-schema.json"). Node: ajv with the JSON Schema draft. Ship the schema in the repo so contract changes fail loudly.

• Faker/Java Faker for random values
• Builder pattern for domain objects
• API seeding via a Test-Data Service, not UI
• Reset via API teardown or nightly cleanup job

try (Connection c = DriverManager.getConnection(url, u, p);
     PreparedStatement ps = c.prepareStatement("SELECT status FROM orders WHERE id=?")) {
  ps.setString(1, orderId);
  ResultSet r = ps.executeQuery();
  assertTrue(r.next());
  assertEquals("PAID", r.getString("status"));
}

Never commit. Use env vars, GitHub Actions secrets, or Vault. In code: System.getenv("QA_PASSWORD"). Rotate quarterly.

Trigger on push/PR, install deps, run tests, publish Allure report as artifact, post Slack summary. Fail the build on any red test or coverage drop.

Allure HTML for QA, Slack summary with pass/fail counts + failed test list for the team, weekly quality dashboard (pass rate trend, flaky top-10, duration) for management.

Use STAR. Example: race condition between token refresh and concurrent API calls that only surfaced under 4-parallel workers. Detected in nightly CI, patched with mutex, added regression test. Prevented a P1 in production.

Show ROI: manual regression takes 3 days × 2 QAs = 48h/release; automation runs in 25 min unattended. Payback in 4 releases. Bring the number, not the opinion.