Python Backend AI Agent Developer

You will help build the software and training data used to develop and evaluate AI agents. You may work on the connectors track, creating Python backends that replicate SaaS tools, or the task track, creating realistic long-horizon tasks and ways to grade them.

• Build Python backends that replicate tools such as Slack, Linear, Jira, Notion, Gmail, and wikis.
• Extend existing connectors and create new connectors within agreed time frames.
• Test connector behavior against the systems being replicated.
• Study workflows and data to identify opportunities for long-horizon task development.
• Write realistic tasks and check their quality, realism, and correctness.
• Create rubrics that distinguish correct, partly correct, and deficient work.
• Use AI coding assistants during development, quality checks, and validation.
• Take part in calibration, quality reviews, and milestone delivery.

This is a remote, four-week contractor engagement. The listing describes the role as entry level, but applicants must already meet the backend development requirements below.

• Pay: $300 per approved task.
• Time:
  At least 20 hours per week.
• Daily commitment:
  Complete at least one approved task per day.
• Task volume:
  No cap while work remains available.
• Location:
  
  Worldwide and fully remote.
• Language:
  English.
• Experience:
  
  At least three years of backend development experience.
• Python:
  Strong experience with FastAPI, Flask, or Django.
• Backend knowledge:
  Scalable backends, REST APIs, and microservices.
• Tools and practices:
  Git, Docker, pipelines, testing, and code-quality practices.
• AI development:
  Daily experience with Claude Code, Codex, Cursor, or Git Hub Copilot.
• Quality judgment:
  Ability to assess software behavior, task realism, correctness, and quality against defined standards.
• Helpful background:
  Replicating SaaS workflows, designing evaluation rubrics, developing long-horizon tasks, or validating AI-agent behavior.

Open Train is the hiring and contracting organization for this role. AI training work uses human-built examples, tasks, and reviews to help AI systems learn how to perform useful work, while experienced contributors judge whether the results are correct and reliable.