Askuity’s QA Automation Engineer protects a simple promise: suppliers and merchants at The Home Depot make profitable decisions from the numbers the platform shows them — so web and mobile quality is not a support function, it is the product’s credibility. Below is the testing record that maps to the role: test strategies designed and executed against live production software, defect verdicts published with evidence, regression discipline inside shipped products, and real-device testing across platforms.
Read closely, this is a quality-architecture brief inside a data product: own the test automation suite, design test plans that keep delivery quality consistent across web and mobile, and make defects so well documented that resolution is fast. It also asks for judgment the posting can’t list: which surfaces carry the most decision weight, what coverage against acceptance criteria actually means, and how mobile quality survives real devices and OS fragmentation. Every one of those is a call I already make — in public, with receipts.
Define the acceptance criteria and the test data first, apply them identically to every target, and let the results argue. A verdict nobody can re-run is an opinion.
A defect report is a document someone else has to act on. Evidence, reproduction, expected versus actual — written clearly enough that escalation is a decision, not a debate.
Every fixed failure becomes a test that runs after every change. Post-release verification on real hardware, because users should never be the test report.
A scoring rubric built as explicit acceptance criteria, standardized test data applied identically across targets, and 247 scored exchanges executed against 32 live production deployments of enterprise AI hiring software. Verdicts — 3 PASS / 42 PARTIAL / 202 FAIL — published per deployment, with transcripts and evidence handling, so every defect finding can be checked and escalated.
A two-validator verification protocol for AI-generated claims: falsifiable claim atoms, evidence ledgers, independence-grouped validation — including a validator slot deliberately left empty, because a system cannot validate its own output — plus right-of-reply and corrections standards. The Ground News audit (16 claims, two independent validators, cleared with exceptions) is the method running in public; the product page is at claimchecker.teambotics.app.
The retrieval-grounded assistant on nikdesign.ca (Supabase/Postgres, HNSW vector index) produced wrong work-history answers in production. I reproduced them, traced them to retrieval settings and grounding rules, hardened the pipeline, and locked the fix in with a regression smoke test that runs the known failure prompts against the live endpoint after every knowledge-base reseed — with deterministic seeding fingerprints keeping the test data repeatable.
A shipped Code2Motion Android build regressed on a physical Galaxy S20; the failure was isolated by controlled comparison against the previous build before users became the test report. Poko ships versioned cross-platform releases — Windows x64 at v0.14.1, other platforms on validated v0.14.0 builds — post-release verification as a standing release discipline across OS and device fragmentation.
takeitandgo runs unattended GitHub Actions discovery, repository verification and production health checks — SHA-pinned actions, minimum permissions, a human approval gate. AgentSurface is a public TypeScript CLI/MCP toolkit; content-buddy’s capture pipeline is Playwright-driven browser automation. Test scripts and tooling in TypeScript and JavaScript, pipelines in GitHub Actions.
Two frontline teams led — computing sales (12) and Geek Squad (6) — in a high-volume retail operation: first in NPS rankings every month from November 2022 to February 2024, and service/CRM workflow redesign that cut repair turnaround from ∼5 business days to under 48 hours. I also piloted BlueBot, a conversational onboarding assistant, on the live floor. Askuity sells analytics into retail; I have worked the floor, the repair operation, and a retail platform team.
A defect report is a document someone else has to act on. Write it like evidence, or it didn’t happen.
Inventory the existing automation suite; map coverage against acceptance criteria for the supplier- and merchant-facing surfaces that carry the most decision weight; publish the baseline quality report.
Automate the highest-weight web and mobile flows, wire real-device execution into CI, and set the defect-documentation standard the whole team writes to.
Post-release verification as a standing release gate, quality reporting the product team plans against, and the mobile test-infrastructure roadmap — framework, device strategy, CI integration — agreed with the QA Manager.