The work

This is a build-first role rather than a pure annotation role. You write real software — services, internal tools, front ends, evaluation harnesses — against model APIs that haven't shipped yet, and the software is the instrument. When a tool-calling interface misbehaves, when a long-context request degrades, when generated code compiles but fails under load, your job is to isolate it, reproduce it, and write it up clearly enough for research to act on.

Day-to-day you can expect: standing up scaffolding around pre-release endpoints (tool interfaces, telemetry, harnesses), prototyping against requirements that change mid-sprint, and delivering working increments without a full spec. You report to the AI lab's engineering manager and coordinate with other engineers on architecture, code quality, and documentation. The cohort is deliberately staffed across language ecosystems — Python, Java, Rust, C#, C++, TypeScript — so depth in one plus working fluency in a second matters more than breadth in all.

What the screen looks for

  • Verifiable production history. 3+ years shipping software at a recognized organization, with specifics: what you owned, what scale, what broke.
  • Genuine full-stack coverage — backend APIs, a modern front-end framework, relational or document data stores, and cloud deployment you actually operated.
  • Cold-start speed. Concrete evidence you've landed meaningful code in an unfamiliar codebase within days, not months.
  • Written clarity. Expect to be asked to explain a technical decision in prose; vague answers under follow-up are the most common failure mode.
  • Career progression that reads as real, not lateral.

Logistics

Full-time, 40 hours/week, weekday-aligned with the client team — this is not a nights-and-weekends side engagement. Remote, with synchronous collaboration expected. Employment is W-2 through Cincinnatus LLC as employer of record (payroll, benefits, compliance), with placement into the AI lab's extended workforce. Pay in the $50–65/hr range has been observed for this listing; actual offers depend on level, language ecosystem, and the lab's requisition, and are set by the platform rather than guaranteed here.