– Data analyst instincts + software engineer’s hands. ~70% of the job is finding a cohort in BigQuery, working out why the data is wrong, and fixing it at scale. Strong SQL and genuine curiosity about data are non-negotiable.

– JVM backend (Kotlin/Java, Spring Boot). Must be able to build their own remediation tooling and checks, not just run queries. Beam/Dataflow, Kafka, Snowflake, Bigtable are a plus, not a must.

– Data quality/observability mindset. Has built monitors, data tests, or anomaly detection. Understands the difference between firefighting and building prevention and wants the second.

– Ownership and autonomy. The backlog is unassigned and partly undefined. They must take a vague ticket, scope it themselves, deliver, and propose what’s next. Fluent English, direct work with the US team, remote.

– Comfortable with legacy and ambiguity. A live, latency-sensitive system where fixes run offline, and validation is statistical, not a green test. No greenfield here.