Speech / Applied ML Engineer

Role overview Own applied machine learning improvements for speech recognition across Southeast Asian languages, accents, code-switching, and noisy audio. The work spans experiments, evaluation, and production integration, with decisions shaped by accuracy, robustness, latency, memory, cost, and reliability.

Responsibilities - Experiment with and tune speech recognition models for multilingual and regional speech use cases. - Design evaluations that reflect real operating constraints, then track performance with datasets and repeatable test suites. - Investigate failure cases in imperfect, multilingual audio data and develop ways to reduce them. - Improve inference speed and resource use, including GPU utilization. - Work with engineering colleagues to integrate model changes into production pipelines. - Record experimental methods, results, and tradeoffs, including when an approach does not work.

Requirements - Experience training or fine-tuning machine learning models and evaluating their performance. - Familiarity with speech or automatic speech recognition systems and their evaluation metrics. - Practical understanding of GPU-based inference and optimization workflows. - Ability to carry work from a hypothesis through investigation and deployment, with attention to production behavior. - Clear communication about evidence, uncertainty, risks, and model failures. - Comfort working with evolving criteria, incomplete signals, and messy real-world data.

Nice to have - Experience with multilingual, Southeast Asian, or low-resource speech. - Exposure to on-device or low-latency inference. - Previous experience deploying machine learning models into production systems.

Benefits and work setup The role offers hands-on experience with production machine learning, collaboration with founders and senior engineers, and opportunities to build a portfolio of evaluated and deployed speech improvements.