AI Engineer

What you'll do

Despite the generic title, this is a specific job: building reinforcement learning environments for coding agents that work through Model Context Protocol (MCP) tools. Each environment is a software engineering problem (a bug, a missing feature, a refactor, a performance issue) that the agent can only solve by pulling information from real MCP servers.

You deliver three things per environment:

• A reproducible setup that behaves the same every run
• Deterministic verification that decides pass or fail without judgment calls
• A golden reference solution

The hard part is the verifier. It has to reward both good tool use and correct engineering, and reject answers that pass by accident. People who have written serious test harnesses or CI gates will recognise the problem.

Who fits

Stated as required:

• Proficiency in one of C++, Python, Java, Go, TypeScript or Rust
• Solid algorithms, data structures and performance tuning
• Experience debugging complex issues and shipping maintainable fixes
• Feature development and refactoring on real codebases
• Clear written communication and remote teamwork

Large distributed codebases, code review culture and some ML familiarity are listed as pluses. No degree and no prior AI experience is asked for.

What it pays

$60–120/hour as the published band, but compensation is output-based: you are paid per task that meets the specification, and a weekly submission minimum applies (the ad leaves the actual number blank). Your effective rate therefore depends on how fast you can build environments that pass review.

This is the same work, band and wording as micro1's MCP Expert listing, posted under a broader title with 100 openings. If you applied there, check with micro1 before applying twice.

How you get hired

Screening questions, a roughly 30-minute AI interview, a technical assessment marked tentative, then hiring manager review. micro1 says roles typically fill within 48 hours and expects first tasks within 24 to 48 hours of onboarding.

Worth knowing

Good:

• Real engineering work with a clear quality bar
• Choose your own hours and days, weekends included
• 100 openings and no degree requirement
• MCP experience is still rare, which helps you stand out

Less good:

• Per-task pay against an hourly band; slow environments cut your effective rate
• A weekly minimum exists but its number is missing from the ad
• The assessment stages are unpaid
• Contractor engagement: no benefits, no notice period, your own tax to manage

About this listing

Posted by micro1 on its own jobs portal about a month before we read it on 24 September 2026, marked high demand. The band, the ~15 hours a week and the hiring steps are micro1's text. Location is "Remote" with no country named. See micro1.