onsite
Deep Learning Engineer - Speech Recognition & Wake Word Detection - 42dot
ML Engineer
Lead the design and deployment of on-device speech models, building STT and wake‑word systems that run efficiently on Linux and Android platforms. Drive research‑to‑service pipelines, optimize models for hardware constraints, and collaborate across AI and automotive teams.
About the role
Key Responsibilities
- Design and train state‑of‑the‑art STT and wake‑word models, leveraging large‑scale speech datasets.
- Build and maintain speech databases, ensuring high‑quality annotations for Korean and other target languages.
- Develop batch and streaming speech applications for Linux and Android, integrating with on‑device inference engines.
- Optimize models for hardware efficiency, applying quantization, pruning, and knowledge distillation techniques.
- Collaborate with research, data, and product teams to iterate on model performance and user experience.
Requirements
- Strong background in deep learning frameworks (PyTorch, TensorFlow) and speech processing.
- Experience with on‑device deployment, model compression, and real‑time inference on mobile/embedded platforms.
- Proficiency in Linux development and Android NDK/Java/Kotlin integration.
- Excellent problem‑solving skills and a passion for cutting‑edge AI research.