Technical Lead QA | Senior SDET | Test Automation Architect
12+ years of experience in software engineering and quality assurance, combining hands-on automation engineering with QA leadership. Experienced in Java, Selenium, Playwright, Appium, Robot Framework, API automation, performance testing, and CI/CD, including quality assurance for GxP-regulated clinical software.
Based in India Previously worked in Germany Open to European relocation
Portfolio projects I built and maintain on GitHub, not company code. Each has CI, a README with setup instructions and a test report; every badge is the live status of the project's workflow.
169 TestNG test executions from 124 test methods, all passing in the scheduled nightly run of 12 Oct 2026: API 57, UI 48, database 13, integration 9, end-to-end 2, framework unit tests 40.
UI, API and database checks of one demo shop kept in the repository; parallel execution, infrastructure-only retries, Allure evidence, Docker + Selenium Grid, nightly CI on Chrome, Firefox and Edge.
47 tests, 95 test executions per full run: 23 API tests once, plus 24 UI tests on each of Chromium, Firefox and WebKit.
Page objects with a header component, typed API clients with zod contracts, login through the API, network mocking; CI runs lint and type checks first, then four parallel jobs.
57 UI, API and network tests, run in CI on Chromium, Firefox and WebKit (one job per browser engine).
Annotation-based page factory, a browser per thread and a context per test, API tests with their own data clean-up, and a trace, video and screenshot kept for every failed UI test.
40 tests on an Android 15 emulator, locally and in GitHub Actions; found 3 real bugs in the demo app, each kept as a test.
Screen objects with built-in waits, device tests, a screenshot and screen recording for every failure, retries only on Android "not responding" dialogs.
30 tests (15 UI, 15 API) in about 70 seconds with 4 parallel pabot processes.
Page objects as resource files, API service objects, Python only where Robot syntax is awkward; CI lints first, then runs on Chrome and Firefox, plus a scheduled nightly run.
46 tests (32 API, 9 UI, 5 hybrid API + UI) in about 1 m 45 s with 4 parallel workers; in CI the UI and hybrid tests run on Chromium, Firefox and WebKit.
A requests API client with pydantic contracts, Playwright page objects, login through the API by token injection, a throwaway user per test, and a screenshot in the HTML report for every failed UI test.
The stress test found the limit of the demo shop: past about 90 orders/s checkout queued without bound (p95 over 3 s). After adding load shedding, the same test stays under 650 ms.
Smoke, load, stress, spike and soak tests of a shop API kept in the repository; SLOs per endpoint as k6 thresholds that fail the CI job, open and closed workload models, and an HTML dashboard per run.
81 Cypress tests: 31 API, 33 UI and 17 component tests; in CI the UI tests run on Chrome, Firefox and Edge.
A React + Node task app kept in the repository, tested three ways: cy.request with zod contracts, page objects with cy.intercept and cy.clock for errors, slow answers and dates, and React components mounted alone.
A TypeScript consumer and a Java Spring Boot provider, 6 interactions verified through a self-hosted Pact Broker; can-i-deploy blocks a breaking consumer change while the provider build stays green.
Consumer-driven contracts with Pact V4: matchers and provider states with parameters on the consumer side, JUnit 5 verification of the real Spring Boot app with pending and WIP pacts, and publish, can-i-deploy and record-deployment in GitHub Actions.
Jenkins configured entirely as code builds a service, deploys it to a kind Kubernetes cluster and runs 21 tests against it in about 30 seconds; GitHub Actions starts a real Jenkins on every push and requires the pipeline to pass.
A declarative Jenkinsfile with parallel stages, parameters, a quality gate and masked credentials; a namespace per build, readiness probes, a smoke Job inside the cluster and API tests through port-forward, all in scripts that CI shares.
Claude drafts Playwright tests from plain-language test cases; a three-stage gate (rules, compile, three runs) decides what a person reviews, and nothing is merged without a named reviewer.
All Java: a demo app in the repository, the Anthropic Java SDK with typed structured outputs, repair rounds fed by the gate, an approve step that marks AI-written tests, and AI failure triage into product bug, test bug, environment or flaky. Answers are recorded, so CI replays them without an API key.
32 hand-written eval cases in six areas, each run three times; prompt-injection and data-leak areas must pass 100%, and the LLM judge is only trusted after it agrees with hand-labelled answers.
An HR policy assistant on Claude and the harness that decides whether it can ship: structured replies, deterministic checks for facts, citations, a canary and colleagues' data, a calibrated groundedness judge, pass rates per area, and 86 tests of the harness itself.
Eight realistic defects are planted in a regressed build of the app, and the suite must report each one with the check named for it: all eight are caught.
WCAG 2.2 AA checks with axe-core in light and dark mode and in every state, keyboard and focus tests for what axe cannot see, reflow and reduced motion, and 21 screenshot baselines rendered in the Playwright Linux image with zero pixel tolerance.
Ten vulnerabilities are planted in a second build of the app: ZAP catches the seven it can see, API tests catch the three access-control flaws no scanner can, and the hardened build must scan clean.
OWASP ZAP baseline and full scans of a hardened and a vulnerable build on a private Docker network, a rules file with FAIL/WARN/IGNORE decisions, a verdict job in CI, and API tests for IDOR, mass assignment and brute force.
Earlier or overlapping frameworks, kept public to show other tools and how my approach evolved.
My 2021 Appium + TestNG framework, upgraded in 2026 to Appium 3, Java client 11 and Selenium 4 without changing its design: one set of tests for Android and iOS through a platform layer, capabilities per platform in properties files, and a custom HTML report with failure screenshots. Runs green on an Android 15 emulator; the iOS side is ready but not yet run.
My January 2024 Selenium + TestNG page object framework for OrangeHRM, upgraded step by step in 2026: explicit waits instead of fixed sleeps, browser and headless mode from a config file, and an Extent report with a screenshot on every failure. The upgrade also fixed a locator bug that had failed the logout test since 2024.
My first Cucumber project from 2022, which no longer built, brought to Java 17, Cucumber 8 (JUnit Platform runners) and Selenium 4 step by step, then split into layers: runners, step definitions, page objects, hooks and utils. Hooks attach a screenshot on failure, and a Scenario Outline that used to pass without checking anything now asserts each row.
148 runnable tests, one per Selenium, TestNG and framework-design topic, plus a PDF of 100+ interview questions.
8 BDD scenarios with Cucumber, Selenium 4 and JUnit, and a custom self-contained HTML report.
A layered REST Assured framework (one manager per API resource) with one end-to-end scenario and a custom HTML report.
An earlier REST Assured framework with Extent, Allure and custom TestNG reports on the same suite.
The original version of the API framework, kept to show how it evolved; Hamcrest matchers and an Extent report.
Employers, titles and dates as on my CV and LinkedIn.
Sycamore Informatics Inc. (employer) · Bengaluru, India · Mar 2025 – present
Leadership
Hands-on engineering
Ephicacy (employer) · Bengaluru, India · Jun 2024 – Feb 2025 · client engagement: Sycamore Informatics
MIVO Mitarbeitervorteile GmbH · Frankfurt, Germany · May 2022 – October 2023
Note: Employment dates reflect the respective company records. The overlapping period occurred during my transition to a role in Germany.
bebo Technologies Pvt. Ltd. · Chandigarh, India · March 2017 – September 2023
Associate Software Architect (Mar 2023 – Sep 2023) · Principal Software Engineer (Mar 2021 – Mar 2023) · Senior Software Engineer (Mar 2019 – Mar 2021) · Software Engineer (Mar 2017 – Mar 2019)
Simplified Fintech Pvt Ltd · Panchkula, India · Apr 2016 – Feb 2017
SSV Technologies and Consultancy Services · Indore, India · Nov 2014 – Feb 2016
Ypsilon IT Solutions Pvt Ltd · Indore, India · May 2014 – Oct 2014
The same list as on my CV.
Selenium WebDriver, Playwright, Robot Framework, Cypress, Appium, TestNG, JUnit, Cucumber (BDD)
Java, Python, JavaScript, TypeScript, SQL
REST Assured, Postman
rfswarm, load and multi-user performance testing, release-to-release response-time baselines
Page Object Model, data-driven testing, keyword-driven testing, BDD, parallel execution
Jenkins, Azure DevOps, Git, GitHub, Jira, TestRail; in portfolio projects also GitHub Actions, CircleCI, Docker, Selenium Grid, Allure, Maven
Test strategy, test planning, RTM, release sign-off, defect triage and management, mentoring, offshore team coordination
GxP, ICH-GCP
Claude for test design, script generation and test data, with human review of all output
M.Tech, Computer Science and Engineering
Galgotias University · 2012 – 2014
B.Tech, Information Technology
Graphic Era University, Dehradun · 2008 – 2012
Hindi (native) · English (professional working proficiency)
Based in India. My earlier work in Germany does not give me a current right to work there or elsewhere in Europe: I would need a work permit for the destination country, and I am open to relocating.
niranjankagri@gmail.com
