Senior Data Architect

We are seeking an experienced Senior Data Architect to design the data foundations that power Oura's AI agents and agentic workflows as part of our unified data mesh platform. Reporting to the Sr. Director of Data Management, this role goes beyond making data queryable by an LLM — it's about making Oura's metrics, entities, and business logic something autonomous agents can reliably monitor, reason about, and act on, within governed boundaries.

We are looking for a Data Architect with deep expertise in modern cloud architectures, semantic modeling, and agent-ready data design — data structured not just for human dashboards but for autonomous decision loops: agents that watch a metric, detect drift, and either recommend or trigger a corrective action. You will bridge business requirements and technical design, acting as the primary blueprint designer for how AI systems and agents access, reason about, and act on Oura's data.

• Architect for Agents: Design data domains, semantic layers, and metric definitions that are directly consumable and actionable by autonomous agents, not just human analysts or single LLM calls.
• Governed KPI & Metric Contracts: Build a governed metrics/semantic layer with clear ownership, definitions, and thresholds so that executive-facing agents (e.g. CxO agents monitoring revenue, churn, COGS, or product quality KPIs) can reliably query, detect deviation, and recommend or execute optimization actions within approved guardrails.
• Agentic Operations: Design the data and control-plane architecture behind agentic operations — agents that continuously monitor governed KPIs, flag anomalies, propose interventions, and (where authorized) trigger downstream workflows or corrective actions autonomously.
• Retrieval & Context Architecture: Design vector-based data architectures, embedding pipelines, and RAG patterns where unstructured context is needed, alongside structured/semantic access for everything else agents reason over.
• Agent Tooling & Interfaces: Define data contracts, schemas, and tool/function interfaces (e.g. MCP-style tool definitions) that let agents query, join, act on, and where appropriate write back to data safely and predictably.
• Multi-Agent & Orchestration Design: Architect the data layer to support agent-to-agent coordination and multi-step workflows (planning, tool calls, state/memory, handoff between specialized agents) rather than single-shot LLM lookups.
• Cloud Infrastructure: Build and optimize Oura's Data Lakehouse (Databricks, BigQuery, Snowflake) at Terabyte–Petabyte scale, feeding both analytics and always-on agentic workloads.
• Agentic Governance: Implement federated governance and guardrails for autonomous agent action — scoped permissions, approval gates, row/column-level security, audit trails, and human-in-the-loop escalation paths for higher-stakes decisions — meeting HIPAA/PHI and security requirements.
• Evaluation & Observability: Design monitoring for data quality, retrieval relevance, agent decision/action accuracy, and drift in the KPIs agents are optimizing, including cost and latency tracking for always-on agent workloads.
• Collaborate: Partner with Data Engineering, Data Science, ML/AI Platform, and business domain owners on a unified approach to human analytics and agentic/autonomous consumption of the same underlying data.
• Standardization: Establish frameworks, a governed data dictionary, and semantic/metric layers that let both people and agents self-serve trustworthy data and act on it consistently.

Experience: 8+ years in data architecture or modeling on cloud platforms (AWS, GCP, Databricks, Azure), with hands-on experience designing data systems that feed agentic AI applications — not only single-turn LLM lookups but continuous, tool-using, decision-making agents.

Cloud Platform & Infrastructure

• Multi-cloud expertise (AWS S3/Kinesis/Glue/Athena, GCP BigQuery/Vertex AI, or Azure)
• Modern data warehousing (Snowflake, BigQuery) and Lakehouse architecture serving analytics, human BI, and always-on agent workloads
• Vector infrastructure: standing up and scaling vector databases (Databricks Vector Search, pgvector, Pinecone) as one tool among several agents use, not the whole architecture
• Docker, Pulumi, and workflow engines for complex data/agent pipelines

Data Modeling & Agentic AI Data Strategy

• Data Mesh principles across agent-facing and human-facing consumers
• Semantic/metrics layers (dbt Semantic Layer, Cube, headless BI) that give agents — including executive/CxO-style agents — a governed, unambiguous source of KPI truth to monitor and optimize against
• Design of closed-loop systems: monitor → detect deviation → recommend/act → log outcome, with clear guardrails on what an agent may do autonomously vs. escalate
• Retrieval/RAG design (chunking, embedding, indexing) for the unstructured-context slice of agent workloads
• Agent tool & context design: tool schemas, function-calling interfaces, memory/state, and orchestration across multi-agent workflows (frameworks such as MCP, LangGraph, or similar)
• MDM/RDM so entities and metrics resolve consistently whether queried by a person, a dashboard, or an agent
• Schema design with Iceberg, dbt (bronze/silver/gold) for agent-readable, action-ready data structures

Advanced Analytics & AI Readiness

• Production AI/ML and predictive modeling (Vertex AI, MLOps) feeding agent decision logic, not just reports
• Architecture for LLM components where they're the right tool, plus non-LLM decision logic (rules, optimization, forecasting) where agents need deterministic or auditable behavior
• Direct experience enabling agentic AI — multi-step, tool-using, sometimes action-taking agents — for operational monitoring, executive reporting, and self-serve workflows
• Automated insights: predictive analytics, anomaly detection, and agent-generated recommendations or actions
• Familiarity with evaluation frameworks for retrieval quality, decision/action correctness, and agent task success, and using them to iterate on the underlying data architecture

Governance, Security & Compliance

• Agent guardrails: scoped access, approval workflows, and audit logging for agents that can read, recommend, or act autonomously
• Data residency and privacy standards extended to embeddings, vector stores, and agent memory
• HIPAA/PHI standards, including for data surfaced to or acted on by agents
• Quality assurance via rigorous validation and agent-facing regression testing before autonomous actions ship

Technical Foundations & Tools

• Kafka, Kinesis, Python, Spark, SQL
• Integration via dbt/Fivetran exposed for BI, LLM, and agent consumption alike
• Orchestration (Airflow, Dagster, dbt, Databricks Lakeflow) extended to agent-triggered or agent-in-the-loop workflows
• Observability for high-cost, long-running, or continuously-active agent workloads
• Spark (Databricks) for ETL, embedding generation, and feature pipelines feeding agent decisions

• Hands-on experience with agent orchestration frameworks (LangGraph, AutoGen, MCP-based tools) in a production, action-taking setting — not just chatbots
• Experience designing or supporting executive/operational agents that monitor KPIs and recommend or trigger business actions
• Healthcare, wellness, consumer electronics, wearables, digital health, or subscription industry experience
• Broad software/systems fundamentals across multiple languages
• Experience embedding AI agents/copilots into product, finance, sales, marketing, ops, and CX workflows, including ones with real write/action permissions
• Ability to navigate hypergrowth while managing regulatory constraints on data used and acted on by AI systems
• Experience with Vertex AI, MLOps, MLflow, dbt, MicroStrategy, Tableau, Power BI, ThoughtSpot
• Advanced degree in Engineering, Data Science, Statistics, CS, or related field