Marathi Language QA Tester — On-site AI Evaluation

Localization Academy Inc.; Barcelona; On-site; EUR 35,000 - 52,000; Full time

Job summary

Welocalize is seeking a native-level Marathi speaker for a confidential, large-scale AI and consumer technology programme. This hands-on, on-site role involves evaluating a pre-release AI model in Marathi and testing performance on client devices, including smartphones and smart glasses.

You will shape how well the product works before broad usage. Responsibilities include reviewing model output for accuracy, testing feature parity with English, and documenting issues with clear steps for

Qualifications

• Native-level or fully bilingual Marathi, written and spoken daily use.
• Confidence with a smartphone and laptop and cross-platform testing.
• Sharp eye for quality and objective judgment in assessments.
• Ability to work independently following detailed test scripts in English.
• Experience in software or app testing or linguistic evaluation is a plus.

Responsibilities

• Review pre-release AI model output in Marathi and judge accuracy, naturalness, and suitability.
• Test Marathi against English for feature parity and non-Latin text handling.
• Identify display or localization errors such as dates, currencies, or untranslated strings.
• Document issues with steps, device, locale, and build details; re-test after fixes.
• Offer language-specific navigation and usability insights for the product.
• Work on-site in a private, secure Welocalize office area; no remote work for this programme.

Skills

Marathi language

Attention to detail

Test script following

Explain issues in English

Job description

Welocalize is seeking a native-level Marathi speaker for a confidential, large-scale AI and consumer technology programme. This hands-on, on-site role involves evaluating a pre-release AI model in Marathi and testing performance on client devices, including smartphones and smart glasses.

You will shape how well the product works before broad usage. Responsibilities include reviewing model output for accuracy, testing feature parity with English, and documenting issues with clear steps for