Senior Researcher

permanent

Fully Remote

Only accepting applications from: United States

Design and develop novel streaming ASR architectures, pushing the boundaries of accuracy-latency tradeoffs in production systems.
Research and prototype LLM-assisted speech-to-text, exploring how large language models can enhance streaming speech recognition and understanding.
Develop new algorithms for streaming speaker diarization, contextual biasing, and multilingual speech recognition.
Drive research from initial experimentation through rigorous evaluation to production deployment, working closely with engineering teams.
Conduct systematic experiments on internal and public benchmarks, with careful attention to evaluation methodology and statistical rigor.
Contribute to technical publications and represent AssemblyAI's research at top venues.
Collaborate on research direction and technical strategy, helping shape the roadmap for Speech AI capabilities.

Experience

Strong research background in speech recognition, with deep understanding of classic streaming architectures (RNN-T, CTC) and modern attention mechanisms.
Expertise in LLMs and language modeling, with ability to bridge speech and language model research.
Proficiency in PyTorch and JAX/Flax with ability to move fluidly between frameworks and implement complex architectures from scratch.
Experience with large-scale distributed training and the practical challenges of scaling speech models.
Track record of publications at top venues (ICASSP, Interspeech, NeurIPS, ICML) or equivalent industry impact.
Strong experimental methodology—systematic approach to ablations, careful baseline comparisons, and rigorous evaluation.
Ability to translate research insights into production-ready solutions; comfort working at the interface of research and engineering.
Excellent communication skills—can articulate complex technical ideas clearly and collaborate effectively across teams.