Mid-Level Everything Engineer (Data & Migrations)

The project

We’re working to modernize a mission-critical legacy system that manages foreign military sales cases for the U.S. government. It’s the system of record behind billions of dollars in international security cooperation. Under the hood: four Oracle schemas (three of them mirrored copies kept in sync by triggers over database links), roughly half a million lines of PL/SQL, hundreds of views and triggers, a large Cognos reporting estate, and decades of accumulated production data with every edge case you can imagine and several you can’t. It works. But we can do better for an important national security mission.

Our job is pretty straight forward. Out with the old, in with the new. We’re going to replace the legacy system with a modern web platform that does all the things; from a functional perspective, a technical perspective, and a compliance perspective. And we’re going to make it happen through quick iterations, built on an Agentic AI Software Development Lifecycle.

How we build (read this before applying)

AI-led development isn’t a novelty act, its just how software is built. Full stop. Every engineer on this team works through agentic toolkits, Codex/Claude Code/etc. Every day. On production code. Our engineering harness (versioned rules, skills, and specs that guide both humans and agents) is a first-class product with its own backlog. We measure ourselves on shipped, verified capability, not lines typed.

What that means in practice:

• You’ll decompose problems into specs precise enough that an agent can implement them directly and concretely, and you’ll own everything you merge, because agent output is your output.
• You’ll review rigorously. The bar is production software for a regulated federal financial system: test-driven development, human review on every merge, and a rigorous CI scans.
• You’ll ship constantly. Small reviewable changes, an always-green main branch, an always-on beta environment stakeholders can touch.
• You get to play with all the latest cool tech as part of your job.

If you think of AI-led development as vibe-coding, or some fad you tolerate between “real” coding sessions, this isn’t your team. If you’ve hand rolled enough code to appreciate how Agentic AI outsources all the tedious bits and lets you build anything, let’s talk.

What “Everything Engineer” means

Everybody builds across the whole stack (UI, API, database, infrastructure, pipeline) because our unit of work is a complete vertical slice of business capability, not a layer. You’ll have genuine breadth: you can build the admin UI for your data quality reports, write an API endpoint, and reason about security controls. And you’ll have real depth in your focus area, where you set the technical bar for the team.

Your focus: Data

You’ll own the ground the whole modernization stands on.

The migration pipeline as a product. Repeatable extract, transform, load, and validate machinery that consolidates three mirrored Oracle schemas into one modern PostgreSQL schema (service discriminator plus row-level security replacing physical schema isolation), slice by slice, with per-slice mapping specs reviewed like code.

Proof, not vibes. Row counts, checksums, financial control totals reconciled per country and case, and golden-master behavioral checks (does the migrated case still price identically?). Your reconciliation reports are literally the government’s acceptance evidence.

Legacy archaeology. Reading PL/SQL packages, triggers, and view definitions to determine which of them hide business rules that must survive, and turning what you find into an evidence-cited knowledge base other engineers build from.

The modern data platform. Transactional PostgreSQL schema design (Prisma), CDC into a dedicated analytics store, dimensional models replacing a 300-plus report Cognos estate, and outbound analytical feeds to enterprise data platforms.

Who thrives here

Deep, wide technical background. You’ve written a lot of code for web-based systems, and you understand the whole ecosystem: cloud infrastructure, CI/CD, security, not just the database. When the pipeline breaks two layers above your SQL, you follow it up.
Agent-native. Real, hands-on mileage with Codex and/or Claude Code (or equivalent) on codebases that matter. You have opinions about specs, context, and harness design based on your experiences.
A tinkerer. Intellectually curious, pulls threads, wants to know why that trigger cascades before deciding whether the behavior is a rule or a bug that became load-bearing. You build quality solutions because you actually understand the problem.
Generous. You share information by default: docs, demos, pairing, better harness rules for everyone. You put the success of the team ahead of your individual success.
Good to break bread with. We’re a small team on a challenging mission.

The stack

TypeScript end to end. Node with a REST API. PostgreSQL with Prisma. React front end. Temporal workflows. AWS GovCloud, defined in AWS CDK (TypeScript). GitHub and GitHub Actions CI. Vitest and Playwright. Codex and/or Claude Code (or equivalent) with a versioned team harness. Slack and Jira. Everything gets wrapped in containers and pushed to EKS or ECS. On the legacy side you’ll deal with: Oracle 19c, PL/SQL, converted-Forté Java, Cognos, Control-M.

Requirements

• US citizenship and eligibility for a US SECRET clearance
• 3+ years of work experience with upward trajectory 
• Strong track record with production relational data systems (schema design, migration, performance) plus real web-stack breadth
• Ability to read and reason about PL/SQL (you don’t need to love it, you need to interrogate it)
• Hands-on experience with agentic AI development tools (Codex and/or Claude Code, or equivalent) on real codebases
• Travel no more than once per quarter

Desired Experience

Heterogeneous migration experience, ideally Oracle to Postgres, with validation and reconciliation rigor; data modeling for analytics; data-quality tooling.

Nice to have: prior work in a regulated or government environment; financial data and audit exposure; synthetic data generation; Cognos experience.