Senior ML Engineer

Role overview

Responsibilities

• check_circle Benchmark TTS and ASR models on Arabic test sets (WER, naturalness, dialect coverage)
• check_circle Fine-tune generative architectures for voice cloning and zero-shot speaker adaptation
• check_circle Build and maintain Arabic-specific data pipelines (audio sourcing, diacritization)
• check_circle Optimize inference for production (quantization, KV cache tuning) to ensure low latency
• check_circle Integrate and evaluate full-stack speech-to-speech pipelines

Basic qualifications

• 5+ years of experience in Machine Learning or AI Research
• Strong proficiency in Python, PyTorch, and the HuggingFace ecosystem
• Hands-on experience training/fine-tuning neural models (TTS, ASR, or Audio Codecs)
• Deep understanding of speech architectures (Whisper, Conformer, HiFi-GAN, or Diffusion)
• Experience with audio-specific processing (VAD, Diarization, Neural Vocoders)
• Proven ability to translate research papers into production-ready experiments
• Knowledge of Arabic NLP challenges (Tashkil/diacritization, dialect variations)
• Experience with inference optimization (quantization, streaming, or TensorRT)

Preferred qualifications

• Arabic language proficiency
• Experience with high-performance CUDA kernels or speculative decoding

Benefits

• check_circle Financial stability;
• check_circle Friendly team and a comfortable working environment;
• check_circle Flexible schedule (8 —10 AM start) with the possibility to work assigned hours and/or adjust work schedule as requested by the manager;

About the company