Staff Software Engineer, Data Platform

Harvey•San Francisco, CA

1d•$231,000 - $340,000•Hybrid

About The Position

Harvey is generating far more data than we currently know how to use well. Product telemetry, agent execution traces, model usage, customer engagement, financial and operational systems — the volume and the number of teams who need to work with it are both growing faster than any single team can serve by hand. As one of the first hires on our central data platform team, you'll build the systems that let every team at Harvey work with data confidently and independently. This is a platform charter, not a pipeline queue: you're building the frameworks, tooling, and paved paths that product engineers, data engineers, and analysts all build on, and you're measured by their leverage and general trust in our data systems. The near-term foundation is ingestion and the warehouse — reliable streaming and batch paths into Snowflake, CDC off production systems, orchestration, and schema evolution that absorbs upstream change instead of breaking under it, and factors in the hard data sensitivity requirements our domain requires. From there the charter expands to the rest of what a modern data platform owes its users: transformation and compute frameworks, self-serve tooling so teams can stand up their own pipelines against well-tested primitives, real-time and stream processing for products and internal systems that can't wait for a nightly batch, and the quality, lineage, and governance layers that make the whole thing trustworthy. Handling PII correctly and honoring multi-region data residency aren't nice to have features here — they're constraints the platform has to satisfy by construction, for customers who are among the most security-conscious institutions in the world. You'll sit between Analytics, Data Engineering, product teams, and Infrastructure. Today this work is distributed and improvised. You'll make it a system, set the technical direction, and help build the team around you. This role is based in San Francisco, CA or New York, NY

Requirements

10+ years building and operating production data infrastructure, with ownership of systems other teams depend on
Deep experience with cloud data warehouses — Snowflake strongly preferred (BigQuery, Databricks, or Redshift experience transfers well) — including performance tuning and cost management
Hands-on experience building CDC and streaming pipelines with technologies like Kafka, Debezium, Flink, or Spark Streaming
Experience with managed ingestion tooling (Fivetran, Airbyte, or similar) and clear judgment about when to buy the connector and when to build it
Strong fluency with workflow orchestration — Temporal, Airflow, Dagster, or similar — operated at scale, not just configured
Strong programming skills in Python and advanced SQL
Experience building frameworks or internal tooling that other engineers use, and the product instinct to know when an abstraction is helping versus getting in the way
Practical experience with data quality, observability, and lineage tooling, and with schema evolution in systems that can't afford downtime
Working knowledge of data governance in a regulated environment: PII classification, masking, access control, retention, and data residency
Familiarity with cloud data services (Azure, AWS, GCP), Kubernetes, and infrastructure-as-code (Terraform, Pulumi)
Comfort operating in ambiguity and defining scope where none exists

Nice To Haves

Experience with dbt and a close working relationship with analytics engineering teams
Experience with lakehouse architectures and open table formats (Iceberg, Delta Lake) or query engines like Trino
Experience operating multi-tenant platforms with strict security, compliance, or data residency requirements
Exposure to data infrastructure for AI products
Prior experience as an early or founding data platform hire at a fast-growing company

Responsibilities

Own the data platform's architecture and technical direction — treating data infrastructure as a software product built from reusable frameworks, and making deliberate build-vs-buy tradeoffs as the platform grows
Build and operate the ingestion layer across streaming, batch, CDC, and third-party connectors, including schema evolution that absorbs upstream change safely rather than silently breaking consumers, so onboarding a new source is a paved path instead of a project
Land data into Snowflake with the freshness, completeness, and cost characteristics downstream consumers can plan around, and define a clean handoff for Analytics Engineering
Own the orchestration platform — scheduling, retries, backfills, and dependency management across the full data graph
Build the transformation and compute frameworks teams can use to process data at scale, and the self-serve tooling that lets product engineers and analysts stand up their own pipelines against primitives you've already made safe
Design and operate stream processing infrastructure for use cases that can't wait for batch — real-time product features, operational alerting, and near-live reporting
Build the trust layer: quality and observability (freshness, validation, reconciliation, anomaly detection, alerting routed to the right owner) alongside lineage, cataloging, and discovery, so anyone can find data and know where it came from and what depends on it
Build the patterns and tooling for PII and sensitive data — classification, masking, retention, access control — and for multi-region residency requirements
Set the technical bar for data at Harvey through design reviews, standards, documentation, and mentorship as the team grows