Quality is an engineering discipline. I build it like one.

Seven-plus years as a quality engineer across AdTech, secure enterprise collaboration, and workplace platforms — on systems ranging from legacy monoliths to event-driven microservices. I design and own test automation architecture: the frameworks, contracts, pipelines, and conventions that let teams ship quickly because of their tests, not in spite of them. I work from Ujjain, Madhya Pradesh — a city that has stayed continuously alive for a couple of thousand years, which keeps my definition of "legacy system" humble.

# the parts a test report can't capture
name: Prayag Vyas
base: Ujjain, Madhya Pradesh, India      # UTC+05:30
hometown_uptime: ~2000 years and counting # older than any codebase I've modernized

off_hours:
  chess:       opening prep, middlegame chaos, endgame discipline — a release cycle in 64 squares
  video_games: field research into how other teams ship (and patch) software
  books:       history and engineering, plus whatever the last bookshop insisted on
  travel:      collecting timezones, not souvenirs
  geopolitics: distributed systems where every node claims sovereignty

warranty: none — but expect strong opinions about flaky tests

Five frameworks, five disciplines. Each is a complete, reviewable system with CI gates, live reports, and a documented reviewer path — I'd rather show evidence than adjectives.

This card is not a third-party stats widget. It reads the same GraphQL contributions API that draws the green graph below, and it is rendered by a small Python generator with its own unit tests, run daily by a GitHub Action that refuses to publish if the tests fail. A profile README is still software.

• Automation is software. Frameworks get architecture, code review, refactoring, and deprecation plans — treat them as scripts and they rot.
• Test at the lowest layer that proves the behavior. UI checks are the last resort, not the default; most business logic wants a service-level answer.
• Optimize confidence per CI minute, not coverage percentage. Every test must earn its execution cost.
• A failing test should explain itself. Diagnostics, redaction-aware logging, and readable reports are features, not afterthoughts.

• Introduced Pact consumer/provider contract testing into a shared automation framework — consumer generation, provider verification, CI integration, contract versioning.
• Built reusable event-driven testing support: producers, consumers, Event Hub workflows, asynchronous assertions.
• Owned and modernized several framework generations: Java · Selenium · TestNG, Selenide · Spring Boot, Python · PyTest, Appium · Experitest.
• Shipped internal productivity tooling: a UUID ↔ legacy-endian converter for MongoDB validation, a standalone ~200-request API regression tool, automated test-data cleanup, a Microsoft Graph auth provider.
• Wired automation into Jenkins, Azure DevOps, Codefresh, and GitHub Actions — parallel execution, faster feedback, diagnostics-first debugging.

Senior IC roles where framework architecture meets technical leadership: Staff SDET · Lead SDET · QA Architect · Automation Architect. Currently deepening distributed systems, Kubernetes, observability, and resilience testing — and always up for a conversation that starts with test strategy and ends somewhere in geopolitics.

Reviewer notes. If you're evaluating me for a role: start with the Verity reviewer guide — it offers 10-minute, 30-minute, and deep-review paths. Then open a live Allure report or the k6 dashboards. Everything above links to something you can read, run, or rerun. If you'd rather talk chess openings or supply-chain chokepoints first, that works too — the inbox is open.