ASR & AI Transcription Specialist (English- typing speed 65+)

Read JD carefully before applying

Title: ASR & AI Transcription Specialist.

Experience: 2+ years

Typing speed: 65+ 99% accuracy

Location: Noida, sector 62 (onsite)

Language: English only (C1 proficiency)

We are staffing a 50-seat specialist bench dedicated to high-fidelity Automated Speech Recognition (ASR) benchmarking, long-transcript auditing, and ground-truth text extraction. In this role, you will audit raw model transcriptions against strict Word Error Rate (WER) thresholds, correct machine hallucinations, perform precise speaker diarization, and resolve complex overlapping dialogue. This role serves as the critical ground-truth verification layer for multimodal speech-to-text models, OCR extraction pipelines, and narrative questionanswering systems

KEY RESPONSIBILITIES & CORE WORKFLOWS

• Long-Form ASR Review: Review and rectify machine-generated transcripts across diverse audio qualities, background noise profiles, and nonstandard acoustics. WER Reduction & Error Tagging: Identify, categorize, and correct substitution, insertion, and deletion errors to drive models toward sub-3% Word Error Rates.

• Speaker Diarization & Audio Segmentation: Apply precise speaker labeling, handle rapid conversational turn-taking, and accurately capture overlapping speech and cross-talk.

• Disfluency & Paralinguistic Formatting: Accurately capture false starts, stutters, filler words, vocalized pauses, and non-verbal acoustic markers based on verbatim protocols.

• Style Guide & Convention Adherence: Switch seamlessly between strict verbatim and clean-read conventions across specialized domain glossaries.

• OCR & Narrative QA Support: Cross-reference extracted text from unstructured document images and video captions to ensure complete semantic and character fidelity