Software Engineer, Data Platform

Numeric•San Francisco, CA

2d•$125,000 - $225,000•Onsite

About The Position

Numeric is building a modern platform for accounting and finance, addressing the challenges of fragmented and complex financial data. The company aims to provide live, rich, and trustworthy data and insights for decision-making. The Data Infrastructure team is central to this mission, building the financial data plane that includes a lakehouse for ingesting, transforming, and serving financial data from various sources. This role is expected to be a key area of investment, with engineers owning end-to-end slices of the architecture.

Requirements

3+ years of software engineering experience, with meaningful time on data-intensive systems.
Strong SQL and a working understanding of columnar stores, partitioning, and query performance.
Experience with a modern transformation / orchestration stack (SQLMesh, dbt, Dagster, Airflow, or similar).
Strong general-purpose programming (we’re primarily TypeScript, with Python mixed in).
Treat auditability and “fail early, fail explicitly” as design constraints, not afterthoughts.
Are comfortable across the stack: TypeScript services, SQL (ClickHouse and Postgres), and transformation tooling like SQLMesh or dbt.
Like being close to the consumers of your data — accounting module engineers, the reporting team, and agents — and shaping the interfaces they use.
Want to learn how accountants think about source data, completeness, and close, and let that shape what you build.

Nice To Haves

ClickHouse, Iceberg, or Athena experience.
Multi-tenant data isolation, PII handling, or SOC 2 / audit-driven controls.
Prior exposure to financial or accounting data (ERPs, GL, subledgers).
Experience with orchestration systems like Temporal or Inngest

Responsibilities

Ingestion at scale: incremental, watermark-based extraction from API sources (NetSuite, Stripe, Brex, Ramp, HubSpot…), file-based S3 inboxes, and Postgres → ClickHouse replication — and making each new source cheap to add.
Data contracts & typed source data: stop downstream consumers from touching untyped raw JSON; enforce schemas at the boundary and make datasets discoverable.
Data Lakehouse schema management: a single declare → codegen → apply pipeline so schema intent is codified, reviewable, and applied consistently across tenants.
Data Security: Scoping queries, building ACLs, and ensuring that we never, ever compromise our customer’s trust
Batch processing & event-driven accounting: decouple event processing from downstream failures; give operators clear “what ran, what failed, and why” answers.
Data correctness & observability: completeness checks rooted in platform data, ingestion health, two-phase re-ingest, and explanatory links from ledger entries back to raw events.
Reporting scalability: the query patterns and materializations that let reporting drill from a board-level P&L to a single transaction without falling over.