Senior AI Engineer – Arabic Speech Technologies

About the role

We are looking for a senior AI Engineer to lead the development of cutting‑edge Arabic speech solutions. You will work on a full‑stack voice infrastructure that includes text‑to‑speech, automatic speech recognition, and speech‑to‑speech conversational systems for multiple Arabic dialects.

Key responsibilities

• Benchmark TTS and ASR models on Arabic datasets, reporting WER, naturalness, and dialect coverage.
• Fine‑tune generative architectures for voice cloning and zero‑shot speaker adaptation.
• Design, maintain, and improve Arabic‑specific data pipelines (audio sourcing, transcription, diacritization).
• Optimize model inference for production using quantization, KV‑cache tuning, and latency reduction.
• Integrate, test, and evaluate end‑to‑end speech‑to‑speech conversational pipelines.
• Conduct research experiments and turn academic advances into production‑ready features.
• Collaborate with cross‑functional teams to deliver high‑performance AI voice products.

Required profile

• 5+ years of experience in Machine Learning, Deep Learning, or AI research.
• Native Arabic speaker with at least B1 English proficiency.
• Proven ability to implement and validate ideas from research papers.
• Experience deploying speech AI systems in production environments.

Required skills

• Python programming.
• PyTorch and Hugging Face ecosystem.
• Training and fine‑tuning neural speech models (TTS, ASR, audio codecs).
• Familiarity with Whisper, Conformer, HiFi‑GAN, diffusion‑based models.
• Audio processing techniques: VAD, speaker diarization, neural vocoders.
• Inference optimization: quantization, streaming inference, TensorRT.
• CUDA kernel development (nice to have).
• Speculative decoding for large generative models (nice to have).