Machine Learning Scientist (Speech AI)

Location: Remote within the United States

Company

hirify.global builds voice AI and text-to-speech models for high-volume enterprise customer experiences, using a proprietary studio-quality conversational speech corpus.

What you will do

• Design, train, and evaluate autoregressive and non-autoregressive speech synthesis models.
• Research full-duplex and half-duplex multimodal architectures, including unified speech-to-speech systems.
• Develop and iterate on speech representations such as neural codecs, semantic tokens, mel features, and continuous latents.
• Build objective and perceptual evaluations focused on speech quality and prosodic control.
• Collaborate with linguists on text-to-speech frontend behavior, including pronunciation and prosody.
• Build data and training pipelines and help move research models toward production.

Requirements

• Deep familiarity with speech synthesis research, including Tacotron, FastSpeech, VITS, VALL-E, and codec-language-model approaches.
• Hands-on experience training neural codecs such as EnCodec, DAC, or Mimi and working with multiple speech representation choices.
• Experience with full- or half-duplex multimodal modeling, streaming speech-to-speech systems, or related architectures.
• Strong attention to data quality, annotation pipelines, and evaluation-set integrity.
• Working knowledge of TTS frontend technologies, including grapheme-to-phoneme conversion, normalization, and prosody.
• Strong PyTorch fundamentals, including training loops, distributed training, and model internals; PhD or equivalent research experience in speech, audio, machine learning, or computational linguistics is expected, unless offset by a strong research track record.

Nice to have

• Multilingual TTS experience.
• Background in prosody or paralinguistics.
• Published work in speech, audio, or core machine learning venues.
• Experience with quantization, distillation, or streaming inference in production.

Culture & Benefits

• Remote-friendly work environment.
• Visa sponsorship available.
• Access to a proprietary full-duplex, studio-quality conversational speech corpus.
• Compute and tooling for research and development.
• Competitive base compensation with meaningful early-stage equity.
• High ownership, strong standards, direct founder collaboration, and low bureaucracy.