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 testsFive frameworks, five disciplines. Each has CI and a documented reviewer path; the available reports and verified evidence differ by repository. Check the linked run and evidence status before treating a capability as verified.
Evidence snapshot (2026-10-17): durable public execution bundles have been verified for ARIA, k6, Playwright, and Selenium. Verity's report page is available, but its required dependency scans still fail on unresolved findings, so Verity has no verified durable bundle in that set. The portfolio dashboard shows the current source-run and freshness status; a reachable report page alone does not mean its security gate passed.
| Framework | Discipline | What it demonstrates | Evidence |
|---|---|---|---|
| verity-policy-coverage-eval-framework | LLM evaluation | Multi-tier evals for a RAG + tool-use assistant: hermetic PR gate, semantic evals, adversarial red-team, judge calibration, mutation testing | Reviewer guide |
| playwright-typescript-framework | Web UI + API | Strict TypeScript, typed API clients, Zod contracts, visual baselines, Axe accessibility checks, sharded CI | Live Allure report |
| k6-performance-framework | Performance | Typed k6 workloads, SLO-based gates, reviewed regression baselines, Grafana/InfluxDB observability | Live perf reports |
| aria-api-framework | API + contracts | Java 21, layered services, Pact consumer/provider contracts, JSON-schema assertions, OpenAPI endpoint coverage | CI runs |
| selenium-testng-java-framework | JVM UI | Selenium 4 + TestNG, Docker Grid, typed configuration, redaction-aware Allure diagnostics, multi-browser CI | Live Allure report |
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.
| Area | Tools |
|---|---|
| Languages | Java 21 · TypeScript · Python |
| UI automation | Playwright · Selenium · Selenide · TestNG · component-oriented page models |
| API & contracts | REST-Assured · Pact · WireMock · OpenAPI · Zod · JSON Schema |
| Performance | k6 · Grafana · InfluxDB · SLO-based gates |
| Events & async | Event Hub validation · producer/consumer testing |
| Mobile | Appium · Experitest STA |
| LLM quality | Multi-tier evals · judge calibration · adversarial testing · mutation testing |
| CI/CD | GitHub Actions · Jenkins · Azure DevOps · Codefresh |
| AI-assisted engineering | GitHub Copilot · Claude · ChatGPT · Atlassian Rovo — accelerators, not autopilots |
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.