AI Prosody and Rhythm Modeling Engineer

Ulara AI

Remote · $160k – $200k / year · Full-time · Posted 4 months ago

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

About the role

Text-to-speech has solved intelligibility. The remaining frontier is naturalness — the subtle patterns of stress, rhythm, pacing, and intonation that make a voice sound human rather than robotic. BLEU scores and MOS ratings can measure the gap, but closing it requires a deep understanding of prosody as a linguistic phenomenon, not just a signal processing problem. Our TTS system powers voice interfaces for 90 enterprise clients, including automotive infotainment systems, healthcare communication tools, and audiobook production platforms. The current system scores 4.1 MOS; our target is 4.5+. The improvement requires better prosody modeling: sentence-level rhythm prediction, contextual emphasis patterns, and handling of edge cases like sarcasm, lists, and code-switching. You'll lead prosody research and engineering on a team of five. You'll design the prosody prediction models, build the training pipeline from labeled speech data, and validate quality with both automated metrics and human listening tests.

Fully remote.

Responsibilities

Design and implement prosody prediction models for TTS post-processing
Build sentence-level rhythm and emphasis modeling from annotated speech corpora
Handle edge cases: sarcasm, enumeration, code-switching, and domain-specific terminology
Create automated prosody evaluation metrics correlated with human MOS ratings
Run human listening studies to validate model improvements
Publish research findings and contribute to the speech synthesis community

Requirements

5+ years in speech synthesis, prosody modeling, or audio ML
Deep understanding of linguistic prosody theory and acoustic correlates
Strong PyTorch skills and experience with sequence-to-sequence architectures
Experience building and evaluating TTS systems end-to-end
Publications in speech synthesis, prosody, or computational linguistics preferred