Senior Analytics Engineer - Decision Intelligence

Role overview

The role sits within an analytics engineering team building a governed decision layer: a single, trusted set of numbers that humans and AI agents can rely on without re-deriving them. Rather than owning ingestion or pipeline orchestration, the engineer focuses on the semantic and modeling layer above conformed warehouse tables. It is a senior, individual-contributor position on a newly forming team, so the first months involve shaping both the technical direction and the working norms.

Responsibilities

- Build and maintain the conformed gold-layer entities spanning customer, account, project, product, and contract, including the cross-system identity resolution that ties records together when no authoritative key exists between source systems. - Author the semantic models that define core metrics once, so dashboards, notebooks, and AI agents return identical answers; version definitions and retire superseded ones. - Create the agent-facing surface, including tools, permissions, and MCP-style interfaces, that allows analyst agents to query data without producing joins that look plausible but are wrong. - Stand up correctness checks: assertions for grain, uniqueness, and cross-system reconciliation routed to named owners, plus evaluation of agent output against known-correct answers, tracking accuracy, refusal, and confidently-wrong rates separately. - Respond to ad-hoc questions from sales, product, and finance, then convert the answers into reusable models so the next version of the same question is cheaper to answer.

Requirements

- Demonstrated AI-native practice, with AI woven into daily work as a way of working rather than as a convenience feature. - Production-grade SQL and Python, including code that ships on a schedule, has broken in the past, and was caught by self-built instrumentation rather than by stakeholder reports. - Prior ownership of a semantic or metrics layer in tools such as dbt, Cube, or LookML, including the experience of changing a definition others already depended on and communicating what moved. - Modeling judgment over tool knowledge: the ability to reason about grain and keys before writing queries, and to tell a data defect apart from an undecided business definition. - Comfort with work-in-progress infrastructure where there is no ticket queue and no inherited process, and both must be built as the work progresses. - Clear writing, since much of the output is a definition or contract others must trust without re-deriving.

Nice to have

- Background in usage-based or consumption business models, where committed, consumed, invoiced, and recognized revenue are genuinely different numbers. - Experience with real-time billing and consumption systems. - Familiarity with lakehouse architectures such as Iceberg, Athena, or Trino. - Contributions to open-source data or AI projects.