Want to know if this job is worth applying to?
New York, New York, United States
Onsite
At its core, Replay was founded to help founders. We started by supporting startups through shutting down, but we have since expanded into unlocking a new revenue stream for all types of businesses.
In 2025, we had a unique insight: the data every company generates each day through collaboration, communication, and building is some of the most valuable training data in the world. Public and synthetic data can only get frontier models so far, so the next generation of model progress depends on real, proprietary data grounded in how actual businesses operate. We are a primary source of it, partnering directly with the frontier AI labs building what comes next.
We have scaled from $0 to a multi-eight-figure run rate in a matter of months
We have raised from top-tier investors, including Floodgate, Afore, Ludlow, and Hustle Fund
We are small enough that you will carry outsized responsibility and grow as quickly as the company does
You will partner with and build for some of the fastest and most important companies in the world
You will help build a massive, category-defining business from the ground floor
We're hiring a backend engineer to make the pipeline that de-identifies sensitive enterprise data correct, replayable, operable, and safe to change.
This is not a conventional data-platform role where a successful job is enough. A pipeline can finish while silently dropping records, duplicating output, applying stale policy, losing lineage, or producing evidence that cannot establish whether a dataset is safe to release. You will own the backend systems and contracts that make those failure modes visible, preventable, and recoverable.
You will work across asynchronous orchestration, batch workers, queues, object storage, databases, many file formats, model-backed stages, deterministic verification, and human review. The role is backend-focused, but the outcome is a product and delivery promise: the team must know what ran, what changed, what remains uncertain, and what can safely happen next.
Design and ship backend systems for multi-stage, high-volume data processing
Define authoritative, versioned contracts for manifests, artifacts, lineage, and state transitions
Make retries, checkpoints, partial failures, replay, backfills, migrations, and rollbacks safe and understandable
Build independent reconciliation and verification instead of treating job success as proof of correct output
Turn escaped and recurring failures into fixtures, regression coverage, release gates, and durable recovery paths
Expose trustworthy run state and safe controls to the products people use to investigate and release data
Diagnose production behavior across code, queues, stores, artifacts, data formats, and deployed versions
Improve correctness, throughput, and operating leverage without weakening privacy, security, or release confidence
Use AI deeply in development and in bounded verification systems, with explicit evaluation and independent checks
Partner with machine learning, applied science, full-stack product, platform, security, and data engineering teammates
A high-risk pipeline boundary has an explicit contract, independent reconciliation, replayable coverage, and safe recovery
Missing, duplicated, stale, or incompatible work is detected before it becomes a customer delivery
Material pipeline state and release decisions are backed by queryable provenance and audit evidence
Recurring reruns, manual interventions, and diagnosis or recovery time decline
Adjacent engineers can add stages and checks through supported patterns instead of one-off scripts and implicit storage conventions
Pipeline changes can be rolled out, backfilled, quarantined, or reversed without delivery heroics
You have at least three years of professional software engineering experience, including personal ownership of production backend systems
You are strong in asynchronous or distributed systems and can reason precisely about queues, concurrency, state, storage, idempotency, partial failure, and recovery
You have worked on systems where output could be materially wrong even when every service looked healthy
You define invariants and use reconciliation, control totals, diffs, replay, goldens, shadow paths, or independent sources to verify correctness
You can design versioned data and artifact contracts and migrate them safely in a live system
You debug from evidence across system boundaries and turn incidents into durable system improvements
You choose technical work based on operator and customer consequences, not architecture in isolation
You use modern AI engineering tools fluently, verify their output, and know when model-backed checks need deterministic guardrails and human review
You communicate clearly across product, ML, data, platform, security, and customer-facing teams
You want to focus primarily on frontend product development or visual craft
You treat pipeline success, uptime, latency, or a green dashboard as sufficient evidence that the output is correct
You prefer isolated infrastructure work without responsibility for data and delivery consequences
You solve partial failure primarily with retries and manual runbooks
You do not want AI tools to be part of your daily engineering workflow
Experience with large-scale batch processing, workflow orchestration, event-driven systems, or data movement
Experience with schema evolution, manifests, lineage, CDC, migrations, reindexing, or backfills
Experience in payments, ledgers, reconciliation, claims, fraud, identity, search quality, observability, or another domain with delayed or weak ground truth
Experience with sensitive or multi-tenant data, least-privilege systems, auditability, quarantine, and fail-closed release paths
Experience combining deterministic checks, synthetic fixtures, offline replay, model-based judges, and human review
Experience with Python, AWS, Airflow, Batch, SQS, S3, DynamoDB, PostgreSQL, or comparable systems
