About Arctic Engine:

Arctic Engines is an enterprise-grade Al human data operations company specializing in high-quality training data, RLHF, and human feedback pipelines for frontier Al models. We are part of the Apna Group, one of India's fastest-growing unicorns, backed by marquee investors such as Lightspeed, Tiger Global, Insight Partners, Peak XV, and others. With native access to Apna's 60M+ workforce, we deliver high-quality training data at unmatched scale and speed.

Job Summary:

We are looking for skilled Odia Transcribers to support a large-scale conversational voice data project for training AI speech models. You will transcribe natural, unscripted Odia conversations covering topics such as Finance, Travel, Healthcare, Retail, Telecom, and more.

Selection Process

• 2–3 minute Odia transcription skill test
• Evaluation & Shortlisting
• NDA + Onboarding
• Production Assignment

Compensation

Payment is based on approved transcription output, with rates shared during onboarding. High-performing contributors may be considered for ongoing projects and QA/Reviewer opportunities.

Requirements

Key Responsibilities

• Accurately transcribe conversational audio verbatim in Odia script (ଓଡ଼ିଆ).
• Preserve fillers, repetitions, stutters, false starts, and natural speech patterns.
• Handle Odia-English/Hindi code-switching, regional accents, and colloquial speech.
• Identify and tag unintelligible speech, overlapping speech, background noise, and other audio issues.
• Follow project-specific transcription and formatting guidelines.
• Maintain 97%+ transcription accuracy and meet assigned TATs.
• Review and correct work based on QA/reviewer feedback.

Requirements

• Native or near-native proficiency in Odia, with excellent reading and writing skills.
• 1–3+ years of experience in transcription, subtitling, captioning, speech annotation, or similar work preferred.
• Strong Odia typing skills; 40+ WPM preferred.
• Excellent listening skills and attention to detail.
• Comfortable working with natural, unscripted, multi-speaker conversations.