Data Architect

What experience should you have:

• 10+ years of data engineering, data platform, or database architecture experience, with at least 3 years owning the architecture, not just the implementation.
• Production experience with event-sourced systems. You have personally implemented or evolved an event envelope, dealt with the upcaster chain problem, and lived with the consequences of an early schema decision.
• Deep Postgres expertise: schema design, indexing, query plan analysis, RLS. Not "I've used Postgres" but "I know what I'd do differently from the last team."
• Experience with streaming platforms: Kafka, Redpanda, Pulsar, or Kinesis. You understand the difference between a topic, a partition, a consumer group, and a saga, and you've designed for all of them.
• Production analytical data infrastructure experience: warehouse design (Snowflake, BigQuery, Databricks, ClickHouse) or modern OLAP patterns. You can take an event log and produce a queryable shape that analysts and AI systems can actually use.
• Hands-on AI/ML data infrastructure experience: vector databases (pgvector, Pinecone, Weaviate, Qdrant), embedding pipelines, retrieval patterns. You have shipped a RAG system that worked and you know why most of them don't.
• Strong SQL, strong enough to be the person other engineers go to.

Nice to have:

• Drizzle, Prisma, or another modern TypeScript ORM at production depth.
• Workflow engine experience (Temporal, Cadence, Airflow) and the data implications of long-running processes.
• Background in feature stores, ML pipelines, or model evaluation infrastructure.
• Semantic layer experience: dbt, Cube, or Malloy.
• Compliance experience: SOC 2, HIPAA, GDPR, data residency. The audit story is part of the data story.
• Open-source work or public writing on data architecture.

What you’ll be doing in this role:

• The canonical event envelope and schema evolution strategy: how events are versioned, how upcasters chain, how projections rebuild, and how we add fields without breaking history. You write the rules and enforce them in code review.
• The transactional data layer: Postgres schema design, tenant isolation via row-level security, indexing strategy, query patterns, and migration discipline. Every table in a data plane carries a tenant_id and a policy. You make sure of that.
• The read-side architecture: projection design, materialized views, semantic layer for downstream analytics and AI consumption. You decide how we move from event log to queryable shape and how that shape stays cheap to maintain.
• The AI/ML data infrastructure: vector stores, embedding pipelines, retrieval-augmented generation patterns, feature stores where applicable, and the lineage that lets us trust what comes back.
• Data governance: lineage, retention, deletion (right-to-be-forgotten in a multi-tenant event-sourced system is a real problem and you'll solve it), and the audit story for every piece of customer data we touch.
• The operational playbook: projection rebuilds, replay procedures, schema migrations on hot data, vector index maintenance, capacity planning. If the data layer goes sideways at 2am, your runbook is what saves the on-call engineer

Benefits

• Total rewards program designed for your protection, peace of mind, and overall well-being.
• Net-zero cost medical option, company contributions to your HSA, fertility support, fully-paid parental leave, a monthly stipend for your lifestyle spending account, and much more.