• Owning quality for a high-scale, low-latency streaming Customer Data Platform deployed on infrastructure we run ourselves (on-prem, Kubernetes), not managed cloud services.

• The platform ingests from thirty or more source systems at hundreds of thousands of events per second, resolves customer identity, computes customer attributes in real time, and emits governed signals to external destinations.

• You will build and own the automation that proves this works: API and end-to-end test frameworks, streaming and data-pipeline validation, latency and load testing against explicit SLOs, and governance testing that proves consent, data-protection, and tenant-isolation rules actually hold at runtime.

• Testing here is evidence-producing, not exploratory only — test outcomes gate whether an artifact is allowed to go live.

Requirements

• 6+ years in QE, majority in test automation (not manual).

• Ownership of test automation frameworks you built/re-architected — not just tests against someone else's framework (core requirement).

• Strong coding in Python, Java, TypeScript, or equivalent: base classes, utilities, config, reporting, test data management.

• API automation (REST Assured, Karate, pytest, Postman/Newman, Playwright API): auth handling, schema validation, contract testing.

• UI automation (Playwright, Cypress, Selenium): POM architecture, dynamic elements, parallel execution.

• Flaky-test diagnosis and elimination: proper waits, isolation, deterministic setup/teardown, root-cause fixes over blanket retries.

• Performance/load testing (JMeter, k6, Locust, Gatling): scenario design, bottleneck ID, throughput/latency validation.

• Strong SQL for data validation: joins, aggregations, window functions, source-vs-target reconciliation at scale.

• ETL/streaming pipeline testing: schema validation, row/value-level comparison, completeness/duplication checks, sampling strategies.

• Event-driven/message queue testing (Kafka): payload validation, ordering, delivery semantics, idempotency, consumer lag/offsets.

• Stateful stream processing testing (Flink): checkpoint/savepoint recovery, state restoration, late/out-of-order events, backpressure behavior.

• End-to-end latency validation against SLOs (p95/p99), distinguishing in-scope processing from excluded external calls.

• CI/CD test integration: trigger config, stages, artifact/report publishing, failure gates.

• Test data management: setup/cleanup, inter-test dependencies, synthetic/masked datasets.

• Structured, auditable test evidence tied to exact artifact version/config.

• Data governance/privacy testing: consent enforcement, classification, tokenization/masking, verifying no raw identifiers leak downstream.

• Identity resolution/entity matching testing: matching outcomes, merge/split, lifecycle transitions.

• Multi-tenant testing: isolation, RBAC/ABAC, cross-tenant leakage checks.

• Replay/backfill/reconciliation validation without re-triggering side effects.

• Lakehouse data validation (Paimon, Iceberg, Delta via Trino/Spark): stream-vs-table completeness/correctness.

• Failure-injection/resilience testing: node/broker/cache loss, recovery integrity, no dupes/loss.

• Kubernetes-deployed platform testing: namespaced envs, Helm, pod/job lifecycle, log/metric access.

• Observability tooling familiarity (Grafana, Prometheus, OpenSearch, tracing).

• AI/ML/LLM output validation a plus: non-deterministic testing, regression measurement, fabrication detection.

• Rigorous defect discipline: reproducible reports, severity/triage judgment, maintained regression suite.

• Domain plus: CDP/customer 360, telecom/high-volume transactional systems, real-time SLA systems, AdTech/MarTech, analytics/dashboards.

• AI tooling familiarity (Claude, Cursor, Codex)