Machine Learning Scientist

• Design, train, and evaluate speech synthesis models, autoregressive and non-autoregressive.
• Drive research on full-duplex and half-duplex multi-modal architectures, including unified S2S systems.
• Choose and iterate on speech representations: neural codecs, semantic tokens, mel features, continuous latents.
• Build rigorous evaluation, objective and perceptual. Hold the bar on quality and prosodic control.
• Collaborate with our linguists on TTS frontend behavior so modeling and frontend choices reinforce each other.

• Deep familiarity with the speech synthesis literature, contemporary and historical - Tacotron, FastSpeech, VITS, VALL-E, the codec-LM lineage. Opinions on what worked and why.
• Hands-on training with neural codecs (EnCodec, DAC, Mimi, etc.) and multiple representation choices.
• Experience with full- or half-duplex multi-modal modeling (Moshi, LLaMA-Omni, streaming S2S).
• Strong attention to detail on data quality. You notice when an annotation pipeline is silently degrading or when an eval set has leakage.
• Willing to roll up your sleeves on unglamorous data and training work - paired with the agency to build pipelines so the team isn’t stuck doing it by hand.
• Working knowledge of TTS frontend (G2P, normalization, prosody) and experience working with linguists.
• Strong PyTorch fundamentals. Comfortable with training loops, distributed training, model internals.
• PhD or equivalent research experience in speech, audio, ML, or computational linguistics or a track record that makes the credential irrelevant.

Nice to have

• Multilingual TTS experience.
• Background in prosody or paralinguistics.
• Published work in speech, audio, or core ML venues.
• Experience taking research models to production: quantization, distillation, streaming inference.

• Direct collaboration with founders, including a CEO with a Stanford computational linguistics PhD.
• Real impact on company trajectory.
• Meaningful equity upside.
• High ownership, high standards, low bureaucracy.

What We Offer

• Competitive base + meaningful early-stage equity
• Remote-friendly
• Visa sponsorship available
• Access to a proprietary, full-duplex, studio-quality conversational speech corpus
• Compute and tooling to do the work
• Direct influence on the future of voice AI

At Rime, we…

• Are outliers
• Cut through the hype to focus on the craft
• Move fast with agency and freedom
• Maintain a growth mindset, finding joy in the struggle
• Do the right things, knowing that it’ll lead to making money

Rime builds voice AI for enterprises running customer experiences at scale. Our text-to-speech models are purpose-built for high-volume conversational deployments, engineered for the pronunciation accuracy, latency, and deployment flexibility that production environments actually demand.

We started from a different premise than the rest of the field: voice AI isn’t bottlenecked by model architecture. It’s bottlenecked by data. So before we trained a single model, we built our own corpus: full-duplex, studio-quality conversational speech, recorded and annotated by PhD linguists. That’s our moat. It’s also why enterprises pick Rime when pilots need to convert into production.

We’re backed by top-tier investors including Unusual Ventures, and we’ve built a team at the intersection of product, research, and craft. Building voice models is an art. We intend to master it.