Senior Software Engineer

$50–100/hrrate as listed

Build reproducible reinforcement learning environments that test whether AI models can fix bugs, ship features, refactor, and optimize performance in real codebases. Roughly 15 hours a week, remote and output-based, with pay observed in the $50–100/hr range depending on task complexity and throughput.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

What the work involves

You are not writing prompts. You are authoring self-contained software engineering tasks that a model has to actually solve, plus the machinery to grade whether it did. A typical unit of work means selecting or constructing a codebase scenario — a real defect, a missing feature, a performance bottleneck, a legacy module that needs restructuring — then packaging it as a reproducible environment: pinned dependencies, deterministic setup, a failing state, and tests or checks that pass only when the problem is genuinely solved.

Each environment ships with a golden reference solution and written reasoning: why the bug occurs, what the correct fix is, which plausible alternatives are wrong and why. That documentation is a large part of the training value. You will also review peer submissions, which in practice means catching environments that are flaky, trivially gameable, under-specified, or whose tests pass for the wrong reason.

What the platform screens for

micro1's screening is AI-led and leans on verifiable specifics under follow-up. Expect to be asked to name real systems you have debugged, describe the actual failure mode, and defend your diagnosis when the interviewer pushes back. Depth in at least one of Python 3, Java, Rust, Go, C++, or TypeScript matters more than breadth. Strong signals include:

Concrete debugging narratives with mechanism, not vocabulary
Fluency in test design and determinism — how you keep an environment from being flaky
Judgment about what makes a task hard for the right reasons versus artificially hard
Clear written technical reasoning; sloppy documentation fails tasks even when the code is correct

No prior AI or ML experience is expected or advantaged.

Logistics

Fully remote and asynchronous, around 15 hours per week, with weekly minimum submission requirements. Compensation is per accepted task rather than hourly; the $50–100/hr band reflects observed effective rates and moves with how quickly you produce environments that pass review. Hiring moves fast — roles often close within 48 hours, and selected experts are expected to submit first tasks within 24–48 hours of onboarding.

What they look for

Python3
JAVA
Rust
Algorithms
Basics C++
Typescript
bug fixing
feature implementation
codebase refactoring
performance optimization

Before you apply

Would you pass this screen?

Drawn from what this listing actually requires. Answered in your browser — nothing is sent to us, and we do not store your answers.

Have you shipped production code in at least one of these languages: Python3, Java, Rust, Go, C++, or TypeScript?
Production coding experience · required
Have you debugged and fixed complex issues in codebases larger than 10,000 lines, and reviewed or validated code written by other engineers?
Code review and debugging at scale · required
Are you available to commit 12+ hours per week on a flexible, output-based schedule, with the ability to start tasks within 24–48 hours of onboarding?
Remote work availability · required
Have you led or significantly contributed to refactoring legacy code or modernizing an older system in a production environment?
Refactoring experience · preferred

0 of 4 answered

Included with this listing

Interview prep — 11 questions

Matched to what this platform's screening actually measures. Answer each out loud in 60–90 seconds: claim → concrete example → implication.

What it tests: This tests whether you've actually worked at scale with real debugging constraints, not toy problems. The listing emphasizes your ability to 'resolve complex software defects' and 'troubleshoot complex systems'—this question separates someone who has done it from someone who has read about it.
A strong answer: Your answer should name a concrete system you worked on (not pseudocode), describe the debugging methodology you used (e.g., binary search through commits, instrumentation, profiling tools), and quantify the impact (latency reduction, bug severity, affected users). It should show systematic thinking, not trial-and-error.
Where people lose it: Candidates often describe a bug that was simple to fix but took a long time to find, then credit luck rather than method. The micro1 platform is asking you to train AI on *debugging strategy*—your answer must show you have a repeatable, intelligent approach to problem-solving, not that you got lucky once.