Job Description
Test Automation and Data Quality - Enterprise Data Platform
As part of our client's engineering organization, we are looking for Agentic Quality Engineers to own how quality is engineered into the data platform. There is no standalone QA gate and no manual regression phase in our pod model.
The successful candidate designs the test strategy, directs AI agents to generate the tests, reviews them for genuine coverage rather than volume, and keeps the quality signals honest and visible to the pod.
This is not a manual testing position and not a test-management role. Candidates whose experience is predominantly manual execution or defect administration are not a fit. You will write code to the same standard as the engineers around you.
Key Responsibilities:
- Own the test strategy for your pod's slice: what gets tested, at which level, and what evidence a release requires.
- Direct agents to generate unit, integration, contract and regression tests, then review them for real coverage and meaningful assertions.
- Build and maintain the automated suites that run on every pull request and every release.
- Build data quality validation into pipelines, covering schema drift, completeness, freshness and referential integrity.
- Instrument quality signals such as coverage, escaped defects and flakiness, and keep them on the pod board where the pod acts on them.
- Apply risk-based prioritisation to find where the product will actually break, including data failure modes.
- Contribute production code, not only test code.
About You:
The ideal candidate is a quality engineer who thinks like a developer and distrusts green dashboards. They can look at a thousand agent-generated tests and tell you which fifty actually protect the product. They are comfortable deleting tests that prove nothing. They are not a manual tester, a gatekeeper or an owner of a defect backlog. They build the automation that makes a gate unnecessary.
Requirements
- Professional test engineering or SDET experience with strong automation ownership.
- Ability to write and review code to the same standard as the engineers you work alongside, in Python, Java, C# or TypeScript.
- Sound test design judgment: boundary analysis, data-driven testing, contract testing and non-functional testing.
- Experience diagnosing and eliminating flaky tests rather than retrying them.
- Experience testing data pipelines and data products, not only applications.
- Familiarity with data quality tooling such as Great Expectations, dbt tests or equivalent in-house frameworks.
- Confident SQL and the ability to reason about data correctness at volume.
- Experience embedding automated tests in CI pipelines so quality is enforced rather than reported afterwards.
- Experience with test environment and test data management.
- Strong GitHub experience including Actions, repository management, branching strategies, pull request workflows and automation.
- Experience working in Git-based development models across multiple teams.
- Practical experience directing AI coding agents such as Factory.ai or GitHub Copilot to produce production code, and reviewing that output critically rather than accepting it.
- Ability to write precise, testable specifications that an agent can build from without coming back for clarification.
- Sound judgment about where agent autonomy is appropriate and where a human decision is mandatory.
- Comfort being measured on agent autonomy coverage and rework rate alongside conventional delivery measures.
- Preferred: Performance and load testing of data platforms.
- Preferred: Security testing fundamentals and familiarity with SAST and dependency scanning output.
- Preferred: Databricks or Microsoft Fabric testing experience.
- Preferred: Experience establishing a test strategy from scratch in an existing codebase.
- Preferred: Experience in global, distributed engineering organizations.
Benefits
- Pods of five with end-to-end ownership. Your pod builds, tests, releases and runs its own slice of the platform. Nothing is handed to a separate QA or operations function.
- Kanban and flow rather than sprints. Work is pulled, work in progress is limited, and throughput, lead time and cycle time are measured. There is no sprint commitment ceremony and no PI planning.
- A six-phase product lifecycle: Plan, Define, Build, Review, Release and Support.
- Tooling: GitHub for Projects, Issues, Actions and GitOps; Factory.ai and GitHub Copilot for agentic engineering; Nullplatform for environments; Aha for portfolio intake.
- Agent-first delivery. We target 60 to 70 percent agent autonomy on eligible work and rework below 15 percent, which shifts engineering effort from typing code to specifying and reviewing it.