Remote Post-Training Research Engineer - Voice AI

PromptCube; San Francisco (CA); Remote; USD 178,000 - 268,000; Full time

Job summary

Deepgram is seeking a Research Engineer, Post-Training to turn vague customer problems into measurable model improvements. You will design evaluations, curate data, fine-tune models, and ship production fixes to customers within weeks.

You will own problems end to end, expand models to new languages, and improve TTS pronounciation and speaker separation while enabling early access to advanced voice capabilities. This role emphasizes rapid experimentation in a fast-changing AI environment.

Qualifications

• Deep experience training, fine-tuning, and evaluating deep learning models.
• Strong Python and PyTorch skills.
• A track record of shipping machine learning to production.
• A habit of owning problems end to end and learning what is needed along the way.
• Good judgment about which problems matter and when a result is ready to ship.
• No specific degree is required; candidates should demonstrate hard problems solved in past projects, shipped systems, or open-source work.
• Active use and experimentation with advanced AI tools, including integrating AI into workflows and building tools into everyday work; AI application is measured as part of performance.
• Comfort adapting quickly to new AI models and modes of working in a fast-changing environment.

Responsibilities

• Identify where Deepgram's voice models fall short and set targets for measurable improvement.
• Design evaluations that show whether a model is better and explain why.
• Curate and generate training data, then fine-tune models to address customer-facing gaps.
• Build working systems quickly and ship them to production, owning a customer problem until it measurably improves.
• Expand models to new languages and regions.
• Improve text-to-speech pronunciation of names, terms, and numbers, and isolate a speaker's voice from background noise and crosstalk.
• Give customers early access to new voice capabilities such as steerable and cloned voices, use their feedback to guide research methodology, and release new benchmarks for Voice AI use cases.

Skills

Deep learning

Python

PyTorch

Production ML

End-to-end ownership

Job description

Deepgram is seeking a Research Engineer, Post-Training to turn vague customer problems into measurable model improvements. You will design evaluations, curate data, fine-tune models, and ship production fixes to customers within weeks.

You will own problems end to end, expand models to new languages, and improve TTS pronounciation and speaker separation while enabling early access to advanced voice capabilities. This role emphasizes rapid experimentation in a fast-changing AI environment.