Speech Machine Learning Engineer

Role overview Build and operate speech models that help language learners identify pronunciation issues and understand how to improve. The role spans model development, audio and text data pipelines, evaluation, and deployment for real-time and batch scoring.

Responsibilities - Train and fine-tune speech models for phoneme-level pronunciation scoring and accent analysis. - Deploy models through Python-based services and manage inference performance in production. - Create evaluation sets representing learners’ first languages and monitor how reliably scores work across them. - Develop audio and text data workflows for cleaning, labeling, augmentation, and versioning. - Explore ways language and speech generation models could provide coaching alongside assessment. - Own relevant infrastructure, including GPU usage, latency, and operating cost.

Requirements - At least three years of experience training, evaluating, and deploying production machine learning models, ideally for speech or audio. - Strong Python and PyTorch skills. - Experience with automatic speech recognition, forced alignment, or phoneme recognition. - Solid grounding in machine learning, statistics, and signal processing. - Ability to take models from experiments into reliable production systems.

Nice to have - Knowledge of phonetics or experience evaluating accents. - Speech-related research publications or open-source contributions. - Experience optimizing models for low latency or on-device use.

Benefits and work setup - Remote work from anywhere with flexible hours. - Competitive pay based on experience. - Early ownership of work and direct collaboration with company founders. - AI tools provided.