Research Engineer, Voice

📋 Description

• Research, develop, and optimize voice/audio neural models (TTS/ASR).
• Build prod training/inference pipelines for voice models, focusing on latency.
• Run end-to-end experiments: data, design, training, evaluation, ablations.
• Collaborate with ML, product, and infra to deploy voice models in Pi.
• Explore neural audio codecs, diffusion synthesis, streaming, multimodal models.
• Develop evaluation frameworks with perceptual metrics and benchmarks.
• Contribute to Inflection's research culture via publications and reviews.

🎯 Requirements

• 2-5 years in audio, speech, or multimodal ML (research or engineering).
• Strong PyTorch proficiency; experience training/debugging large-scale models on GPUs.
• Solid understanding of audio/speech: spectrograms, mel, vocoders.
• Able to take ideas from prototype to production; CUDA-aware training loops.
• Familiar with diffusion, autoregressive codecs, flow-matching for audio.
• Clear, collaborative communication for cross-functional teams.
• BS/BA in CS/EE/Linguistics or related; MS/PhD preferred.

🎁 Benefits

• Medical, dental, and vision coverage.
• 401k matching.
• Unlimited PTO.
• Parental leave and caregiver flexibility.
• Visa support for international Bay Area employees.