Machine Learning Researcher, Audio

This is a fully remote position, open to applicants in California.

📋 Description

• Design and train large-scale text-to-speech models that produce expressive, controllable, and human-like output.

• Develop TTS architectures based on neural audio codecs for efficient and high-fidelity audio generation.

• Enhance prosody modeling, question inflection, emotional expression, and robustness across multiple speakers.

• Optimize for low-latency, real-time inference in production environments.

• Construct and fine-tune large-scale ASR systems that are resilient to accents, noise, telephony artifacts, and code-switching.

• Utilize self-supervised pretraining and large-scale weak supervision techniques.

• Boost transcription accuracy for real-world enterprise applications, focusing on structured extraction and conversational subtleties.

• Investigate and apply neural audio codecs that achieve high compression with minimal perceptual loss.

• Examine both discrete and continuous latent representations to enable scalable speech modeling.

• Create codec architectures that facilitate downstream generative modeling and controllable synthesis.

• Curate and process extensive audio datasets covering various languages, speakers, and environments.

• Develop staged training curricula and strategies for data filtering.

• Scale training across distributed GPU clusters, emphasizing cost, throughput, and reliability.

• Design ablation studies to isolate the effects of architectural modifications.

• Assess improvements using objective metrics and perceptual evaluations.

• Validate concepts through targeted experiments that confirm or refute hypotheses.

• Collaborate with engineering teams to transition research into production inference systems that handle millions of calls daily.

⛳️ Requirements

• Proven experience with self-supervised learning, multimodal modeling, or generative modeling techniques.

• Capability to devise new formulations and implement them effectively.

• Practical experience in building or scaling TTS, STT, or neural audio codec systems.

• Familiarity with large-scale speech datasets and the variability of real-world audio.

• Strong intuition regarding audio quality, prosody, and the dynamics of conversation.

• Experience in training and deploying large models on contemporary accelerators.

• Knowledge of inference optimization methods, including quantization, kernel optimization, and memory efficiency.

• Understanding of real-time constraints in telephony or streaming contexts.

• A history of designing controlled experiments and significant ablation studies.

• Comfortable working with both offline benchmarks and live production metrics.

• Adaptability to fast-paced startup environments.

• A strong ownership mindset from research through to deployment.

• Experience with large-scale distributed training.

• A PhD in ML, AI, or a related field, or equivalent research impact (bonus points).

• Research publications or contributions to open source projects in the fields of speech or language AI (bonus points).

• A background in real-time speech systems or telephony (bonus points).

🏝️ Benefits

• Healthcare coverage.

• Dental insurance.

• Vision insurance.

• Meaningful equity in a rapidly growing company.

• Access to every tool necessary for success.

• A beautiful office located in Levi's Plaza, SF, featuring rooftop views.