Arabic Speech Recognition (ASR) & NLP Engineer

Must have Qualification:

1. 5+ years in speech technology (ASR/TTS) or Arabic NLP, with production ASR deployment experience.

2. Native or near-native Arabic with strong command of Gulf dialects; fluent English.

3. Hands-on with modern speech stacks (e.g., Whisper-family, wav2vec2, NeMo, Kaldi, or commercial engines) including fine-tuning and evaluation.

4. Streaming/real-time speech processing experience; solid Python and audio DSP fundamentals.

5. Rigorous evaluation practice: WER/CER methodology, test-set construction, error analysis.

Key Responsibility:

• Select, benchmark, fine-tune, and deploy Arabic/English ASR for live sessions: streaming transcription, diarization, and code-switching handling.

• Optimise for Gulf/Emirati Arabic and challenging conditions: children’s and adolescent speech, distance microphones, room reverberation.

• Build the prosody/voice-characteristics pipeline (pitch, energy, pace, pause patterns) as noncontent behavioural signals.

• Own speech-quality measurement: WER per language/dialect, human usability ratings, and transparent reporting.

• Localise the product experience with the UX lead: Arabic transcripts, RTL rendering, terminology consistency.

• Contribute Arabic scenario scripts and act as native-language quality gate through rehearsals and expert reviews.