Senior Backend and Data Engineer - PT - EMEA/LATAM/NA

The role

Standard Data builds a continuous, live flow of verified vertical data and physical-world data (records, imagery and video) for AI agents, robots and the labs and firms behind them. Buyers take a versioned dataset or subscribe to the flow; the first product is a dataset and the flow follows. Everything runs on ADTL, a refining pipeline that ingests raw evidence, normalizes it, resolves entities, validates it with confidence scores and provenance, and structures it. Every value must trace back to its raw source.

You will be the first engineer on the platform. A core pipeline is built (Python, FastAPI, PostgreSQL, Next.js, 12 merged pull requests, CI as a merge gate). The work now is to wire it end to end, generalize it, load real data and put it in front of paying customers through a live API and MCP server. The founder is non-technical and makes the decisions. AI coding agents (Codex and Claude Code) write much of the code under direction. You own the architecture, the review and the deployment.

What you will do

• Finish the pipeline. Review the open pipeline-wiring PR (#13) and finish it so each stage commits independently, and stand up staging and production with CI-gated deploys, backups and monitoring.
• Generalize the data model. Extend entity resolution beyond businesses to sites, projects, asset types, credentials and jurisdictions with dated relationships. Add a fact format that tracks when something was true and when we learned it, with status labels (verified, probable, inferred, not found).
• Build connectors. County and parish records, Secretary of State filings, environmental permits, grid documents, and federal and aviation records, across several states and three project types (data centers, semiconductor fabs, launch sites), each with a stable source ID and a no-cloud option.
• Keep the data flowing. Versioned dataset releases first, then a refresh scheduler, change detection, freshness tracking per source (update cadence and lag) and a subscription feed. Imagery follows soon after the first dataset.
• Ship the customer surfaces. A unified query layer, versioned REST API with hashed per-customer keys, metering and rate limits, and a remote MCP server on the same service layer.
• Add payments. Billing and x402 per-call payments, tested on a test network first.
• Extend to media. Object storage, video handlers and a QC module that scores footage and matches outcome labels. Later: lab export formats, a client-embedded deployment and an air-gapped package.
• Write it down. Design notes for the open questions (deletion across stages, encryption, rights and entitlements), and short written reports with every pull request.

What you bring

Required

• Strong Python and FastAPI experience in production.
• PostgreSQL schema design and migrations (Alembic or equivalent), including JSONB.
• Data pipelines with ingestion, normalization and validation. Entity resolution or record linkage is a strong plus.
• API design with authentication, API keys, rate limiting and usage metering.
• Docker, CI/CD and deploying to a managed host (Railway, Render, Vercel or similar).
• A testing habit: you show commands run and literal inputs and outputs, not "tests pass".
• Comfort directing AI coding agents and reviewing their pull requests critically.
• Clear written English and the judgment to stop and ask when a specification is ambiguous.
  Helpful
• Geospatial and public-records data (PostGIS, parcels, permits), and satellite or aerial imagery pipelines.
• Schedulers, change detection and freshness monitoring for data feeds.
• Media pipelines, object storage, video processing or computer-vision QC.
• MCP servers, x402 or other payment protocols.
• Security work: key management, encryption, air-gapped or on-premises deployments.
• Robotics data formats such as LeRobot or RLDS.

How we work

• One ticket, one branch, one pull request. CI must pass before merge.
• The architecture rules in the developer briefing are fixed: provenance is permanent, config over code, human review is a pipeline state, the core runs offline.
• You size the work yourself and report estimates with confidence levels.
• A weekly review with the founder, on Thursdays.
• You may change the workflow. Keep the verification habit and the stop-and-ask habit.

First weeks

Week 1
Repository state confirmed and test suite run; sizing for every build step; hosting chosen; staging started; MCP authentication and x402 proposals written

By Nov 30 (target, confirmed after sizing)
Pipeline wiring finished; staging running the dashboard; design note for the generalized entity model