Open Source Maintainer

Il y a 2 jours

Paris, Île-de-France Training Turk Télétravail Temps plein 153 000 € - 218 000 €/an
  • AI code critique

$150/hr

Mercor

Hourly, 40 hours a week

Provide expert open‑source code review and feedback to train AI models for software engineering tasks.

What you would do

  • Create challenging tasks and reference solutions based on a mature codebase
  • Evaluate AI‑produced code, tests, and grading rubrics
  • Write precise rationales explaining strengths and deficiencies
  • Build and run the project locally, often with Docker
  • Coordinate with other subject‑matter experts for data consistency

Who they want

  • Proven maintainer or core contributor to a popular repo (10+ PR reviews or 3+ merged PRs recent)
  • Hands‑on ability to build, test, and Dockerize the project
  • Strong written communication for detailed feedback
  • Available 30‑40hours/week, no concurrent AI‑lab work
  • Authorized to work in their country and pass background checks

Main skills

Open‑source maintenance Code review expertise Docker environment setup

What The Interview Asks About

  • Task design from repo

“How would you design a refactoring task for a repository you maintain that targets a known performance bottleneck, ensuring the solution can be automatically graded?”

Why this comes up

Shows they can translate real‑world codebase intricacies into meaningful evaluation tasks, crucial for generating useful training data.

  • AI code critique

“When reviewing AI‑written unit tests for a library you own, what specific signs would indicate the tests miss edge‑case behavior?”

Why this comes up

Demonstrates their capacity to spot subtle flaws in AI‑generated code, directly impacting model improvement.

  • Dockerized build verification

“Describe the steps you would take to containerize a legacy project with multiple dependencies to verify an AI‑suggested patch builds successfully.”

Why this comes up

Validates practical skills to reproduce environments, essential for confirming task correctness.

  • Technical rationale writing

“What structure would you use to write a rationale explaining why an AI‑generated implementation violates the project's coding standards?”

Why this comes up

Assess ability to convey complex judgments clearly, which the AI training pipeline relies on.

  • Collaboration with researchers

“If a researcher proposes a new evaluation metric that conflicts with the repo’s existing style guide, how would you reconcile the two?”

Why this comes up

Ensures they can align engineering feedback with research goals, fostering consistent data quality.

A task you may get

Candidates may be asked to review a pull request from an AI model, run the code in Docker, and write a concise feedback report with a reference solution, or null

How to prepare

  • Clone a popular open‑source project you’ve contributed to and practice building it with Docker
  • Create a mock AI‑generated PR and write a detailed review and rationale
  • Refresh knowledge of code‑review best practices and repository contribution guidelines

The facts

  • Pay $150/hr
  • Hours Hourly, 40 hours a week
  • Where Remote, US, Canada and Europe
  • Open to USA