Automatic Speech Recognition

As an ML Research Engineer with 2+ years of experience, you bridge the gap between cutting-edge research and production systems. You are passionate about training models that excel not only on benchmarks but also in real-world applications. Your enjoyment lies in delving deep into model architectures, experimenting with training techniques, and constructing robust evaluation frameworks to ensure model reliability in critical applications. - Train and fine-tune models for speech recognition and natural language processing in multilingual healthcare contexts - Develop specialized models for domain-specific tasks using fine-tuning and optimization techniques - Design and implement comprehensive evaluation frameworks to gauge model performance across critical metrics - Build data pipelines for collecting, annotating, and augmenting training datasets - Research and implement state-of-the-art techniques from academic papers to enhance model performance - Collaborate with AI engineers to deploy optimized models into production systems - Create synthetic data generation pipelines to tackle data scarcity challenges Qualifications: Required: - 2+ years of experience in ML/DL with a focus on training and fine-tuning production models - Deep expertise in speech recognition systems (ASR) or natural language processing (NLP), including transformer architectures - Proven experience with model training frameworks such as PyTorch, TensorFlow, and distributed training - Strong understanding of evaluation metrics and the ability to design domain-specific benchmarks - Experience with modern speech models like Whisper, Wav2Vec2, Conformer, or LLM fine-tuning techniques such as LoRA, QLoRA, full fine-tuning - Proficiency in handling multilingual datasets and cross-lingual transfer learning - Track record of enhancing model performance through data engineering and augmentation strategies Nice to Have: - Published research or significant contributions to open-source ML projects - Experience with model optimization techniques like quantization, distillation, pruning - Background in low-resource language modeling - Experience in building evaluation frameworks for production ML systems,