Brazil & Argentina Software Engineers: Coding Tasks for AI Evaluation

What you'll do

The study checks whether coding tasks built to evaluate AI agents are realistic, accurate and well calibrated. You review a series of proposed tasks, judge whether they suit testing an agent, and verify the logic of the evaluation harnesses behind them. The session is live: you share your screen, walk through the environments, and point out flaws or improvements. Expect a mix of code review, technical discussion and direct feedback on how tasks are structured.

Who fits

• Software engineers based in Brazil or Argentina (a stated location requirement)
• Direct experience building, reviewing or testing evaluation harnesses and programming tasks
• Backend developers, full-stack engineers and QA automation specialists are named as welcome
• The ad does not name languages or programming stacks, and does not say whether the session is conducted in English

What it pays

USD 40 per hour, as listed by Terac. The ad does not say how many hours the session takes or whether more sessions follow.

Worth knowing

Good:

• Paid in US dollars for local-market engineers
• Work that sits close to how AI coding agents are actually tested
• Screen sharing means you can discuss real code, not abstractions

Less good:

• Restricted to two countries, so most engineers cannot apply
• It reads as a session-style study, with no ongoing hours promised
• Harness experience is required, which is narrower than general software skills

About this listing

Terac posted this on 1 October 2026 and lists it until 30 November 2026. The Apply button opens this study on Terac; you sign in or create a Terac account there to apply. Terac's location field shows Worldwide, but the ad text asks for engineers based in Brazil and Argentina, so we treat those as the countries. The comments are ours. See Terac.