qa

🧪 QA live

From test design to SDET and AI-system QA — 10 stages, fast track 12–16 weeks, full track 10–14 months

▸ Stage 0. QA fundamentals and test design (3–5 wk) live

  • Testing principles, levels and kinds of testing
  • Design techniques: equivalence classes, boundaries, decision tables, peer-to-peer
  • Artifacts: test plan, bug report with severity and priority, traceability
  • Requirements, acceptance criteria, BDD, Definition of Done
  • Processes: Agile, Scrum, Kanban, release flows, the QA role
  • Risk-based testing: priority as probability and damage
  • Portfolio QA-1: test design artifacts plus 10 quality bug reports

▸ Stage 1. Technical basics (3–4 wk) live

  • Git and GitHub: branches, pull requests, reading diffs
  • HTTP: methods, codes, headers, cookies, caching, CORS, auth
  • DevTools: Network, Console, Application tabs; curl
  • SQL: SELECT, JOIN, GROUP BY; verifying data after operations
  • Application layers: client, server, API, DB, queues, cache
  • Logs and tracing: reading service logs, searching by request_id

▸ Stage 2. API testing (3–4 wk) live

  • REST semantics, idempotency, pagination, versioning, OpenAPI contract
  • Auth schemes: Bearer, JWT, OAuth 2.0; roles and permissions
  • Postman or Bruno collections; httpx plus pytest in code
  • What to test: negatives, validation, boundary, IDOR, error contracts
  • Consistency with the database; backward compatibility
  • GraphQL and gRPC basics; mocking external services (WireMock, respx)
  • Portfolio QA-2: API suite that catches IDOR and duplicate POSTs

▸ Stage 3. pytest automation framework (4–6 wk) live

  • Fixtures with scopes and cleanup, parametrization, markers, conftest
  • Plugins: pytest-xdist, pytest-timeout, pytest-randomly, pytest-cov, Allure
  • Framework layout: tests per level, clients, environment config
  • AAA or Given–When–Then discipline, one behavior per test
  • Test data: factories (factory_boy, Faker), isolation and cleanup
  • Determinism: frozen time, seeds, mocked boundaries, no sleeps
  • Property-based (Hypothesis) and snapshot tests where they fit

▸ Stage 4. UI automation: Playwright (4–6 wk) live

  • pytest-playwright: page, context, browser fixtures; run options
  • Resilient locators: get_by_role, get_by_label, get_by_test_id
  • Auto-waiting with expect; zero sleeps in E2E suites
  • Isolation: fresh browser context per test; saved storage_state
  • Failure debugging: Trace Viewer, screenshots, video
  • Page Objects and agreed data-testid contract with developers
  • Cross-browser, mobile emulation, visual regression, accessibility
  • Portfolio QA-3: 10–15 E2E scenarios, flaky rate under 1%

▸ Stage 5. CI/CD, environments, data, reports (3–4 wk) live

  • GitHub Actions: matrices, dependency cache, artifacts, services for PostgreSQL
  • Docker Compose for test environments; healthchecks; testcontainers
  • Data factories and seeding; isolation by schema or transaction
  • Parallel runs: pytest-xdist, suite splitting across runners
  • Quality gates: red CI blocks merge; quarantine policy for flaky tests
  • Allure reporting; pipeline metrics: feedback time, flaky rate
  • Portfolio: pipeline lint → unit → API → E2E → report

▸ Stage 6. Non-functional testing (4–6 wk) live

  • Performance: load, stress, spike, soak; RPS, percentile latencies, saturation
  • Open-loop scenarios (Locust, k6) built from real user flows; bottleneck hunt
  • Security per OWASP Top 10 and API Security Top 10; IDOR checks
  • Tooling: ZAP, Burp, pip-audit, SAST, image scanners
  • Accessibility: WCAG 2.2 A/AA, axe-core, keyboard navigation
  • Compatibility and localization: timezones, encodings, RTL
  • Portfolio QA-4: load report plus OWASP-based security report

▸ Stage 7. SDET practices (4–6 wk) live

  • Framework architecture: layers, configurability, docs and onboarding
  • Code testability: DI, ports and adapters, managed time and randomness
  • Contract tests: consumer-driven (Pact), schema-driven (schemathesis)
  • Mutation testing: evaluate your tests, not line coverage
  • DB migration and concurrency suites; idempotency and outbox events
  • Production testing: feature flags, canary, synthetic monitoring
  • Flaky management: rate metric, quarantine, fix-or-delete rule
  • Portfolio QA-5: SDET template framework with gates and docs

▸ Stage 8. QA for AI systems (3–4 wk) live

  • Why LLM QA is not pass/fail: quality criteria, drift, context dependence
  • Evals: golden questions, automated checks, LLM-as-judge calibrated on people
  • RAG metrics: recall@k, MRR, nDCG for retrieval, faithfulness for generation
  • Agent QA: trajectories, tool-call limits, dangerous-action confirmations
  • Security: prompt injection, jailbreak, red teaming (promptfoo, garak)
  • Responsible AI: bias, fairness, transparency, regulation
  • LLM as a QA helper: review everything, never leak secrets

▸ Stage 9. Quality leadership (as you grow) live

  • Quality strategy: risk-based, pyramid levels, release go/no-go criteria
  • Processes: Definition of Ready and Done, gates, bug triage, bug bash
  • Dashboards: escaped defects, feedback time, flaky rate, DORA
  • Communication in risk and cost language for business
  • People: hiring QA, mentoring, guild; shift-left with developers
  • Management tools: TestRail, Zephyr, Allure TestOps — process beats tooling

Source: roadmaps/qa.json · ETA: 2026-11