Full Stack Engineer

Scalable Path is looking for a Senior AI-First Full-Stack Type Script Engineer with Postgres Focus to join the team. This is a remote, full-time (40 hours/week) position.

The client is a well-backed New York startup in building a generic loan origination and automated underwriting platform from scratch.

The client is building a loan origination and automated underwriting platform. The build is AI-primary: most of the code is generated by AI against a loosely specified architecture that needs input/ratification from you. You are the human who owns the data side of that. You direct the build, review what comes out, finish what the AI cannot, and sign off on merges as one of two engineering reviewers.

Your lane is the system of record — an immutable, append-only store of application facts, decisions, and their full evaluation traces, reproducible byte-for-byte months later — plus the intake API that feeds it, the historical loader that backfills it from years of prior applications, and the data side of document extraction. The hard parts are schema design under immutability constraints, access enforced at the database layer, as-of queries over history, getting events out of Postgres reliably — and knowing, when the AI's version is subtly wrong, exactly why.

This suits someone who would rather own outcomes than type every line, who can read a schema or a migration and spot the flaw before it ships, and who finds a codebase built from specs and vectors rather than tribal knowledge a relief. If you have built ledgers, financial calculation engines, or anything where "roughly right" was never acceptable, this will feel familiar.

• Own the system of record: append-only envelope storage for application facts, with access enforced at the database layer (grants + row-level security) rather than in application code.
• Persist decisions and their complete evaluation traces, with snapshot hashing, so any historical decision can be re-derived and verified byte-for-byte.
• Own the headless intake API and programmatic path — how applications enter the platform — contract-first, with the API contract as the artifact.
• Build the event path out of Postgres: transactional outbox into AWS Event Bridge, with the idempotency and delivery guarantees that implies. Design the boundary with the Temporal workflow layer (owned by the platform-focused engineer) so outbox delivery and workflow retry semantics compose rather than collide.
• Design and own the Postgres schema — partitioning, indexing, and query-plan-level performance against append-only tables that only ever grow.
• Own the historical application loader: ingest years of prior applications and their outcomes into the fact model to support backtesting.
• Own the data side of document extraction: fact-model schemas, validation of extracted candidate facts, confidence thresholds and routing into the store.
• Build KPI and reporting surfaces over the system of record — as-of queries, window functions, history that never rewrites itself.

• Senior full-stack Type Script with a database center of gravity. You direct AI tooling to do most of the typing, and you could code all of it yourself if you had to — the second part is what makes the first part safe.
• Deep PostgreSQL — not "I use an ORM" deep. Schema design, row-level security and grant-based access control, window functions, partitioning, and performance work at the query-plan level.
• Production event-driven design: transactional outbox or equivalent, idempotency, at-least-once delivery and what it actually costs you.
• Production AWS. We are AWS-native — Event Bridge, S3, KMS, RDS/Aurora, Bedrock.
• You will work in a codebase that bans floating-point arithmetic and…