US: $174000 - $252000 (USD) + 15% bonus target + equity + benefits
Learn more about benefits at Google.
Define and lead research agendas, experimental designs, and evaluation benchmarks in sequence, speech, audio, and multimodal modeling.
Co-design and optimize models for scaling laws, memory or cache efficiency, distillation, and low-latency streaming inference in JAX, PyTorch, or TensorFlow.
Contribute to open-source toolkits, and mentor researchers.
Minimum qualifications
PhD in Computer Science, Linguistics, Electrical Engineering, Mathematics, Physics, a related field, or equivalent practical experience.
2 years of experience leading a research agenda in sequence, speech, or audio modeling.
Experience with sequence-to-sequence models across state-space or linear attention, token-free latent representations, speech-text modeling, diffusion, flow-matching, speculative decoding, or streaming alignment.
Experience with hardware-aware model co-design, scaling laws, memory or cache optimization, or streaming inference latency in JAX, PyTorch, or TensorFlow.
One or more peer-reviewed publications in machine learning, speech, or audio conferences or journals.
Preferred qualifications
10 years of experience in the theoretical and algorithmic aspects of sequence modeling and machine learning research.
4 years of experience setting research agendas across multiple projects or teams, and publications in NeurIPS, ICML, ICLR, JMLR, IEEE Transactions, Nature, Interspeech, ICASSP, or ACL.
Experience building, scaling, pre-training, and post-training foundational audio, speech, or multimodal models using reinforcement learning alignment or multi-resolution modeling.
Experience with model compression, distillation, or machine learning acceleration on TPUs or GPUs, and establishing evaluation benchmarks.
Contributions to open-source machine learning or speech toolkits such as JAX, PyTorch, Hugging Face Transformers, Kaldi, K2, or ESPNet.
PhD in Computer Science, Linguistics, Electrical Engineering, Mathematics, Physics, a related field, or equivalent practical experience.
2 years of experience leading a research agenda in sequence, speech, or audio modeling.
Experience with sequence-to-sequence models across state-space or linear attention, token-free latent representations, speech-text modeling, diffusion, flow-matching, speculative decoding, or streaming alignment.
Experience with hardware-aware model co-design, scaling laws, memory or cache optimization, or streaming inference latency in JAX, PyTorch, or TensorFlow.
One or more peer-reviewed publications in machine learning, speech, or audio conferences or journals.