3 months ago
Remote, United States or San Francisco, CA, USASenior / Staff+
H1B sponsor
Base Salary
$150k - $250k/yr
Responsibilities
- Architect horizontally scalable training systems to shorten the STT/TTS model development cycle.
- Build high-throughput pipelines for ingesting, versioning, and preprocessing large unstructured audio and text datasets.
- Own research infrastructure including distributed training, job orchestration, experiment tracking, and storage.
- Ship automated evaluation tooling for rapid research validation.
- Turn research proofs of concept into production-grade systems for Deepgram’s STT and TTS models.
- Build internal tools and dashboards that provide visibility into and control over ML workflows.
- Write design documents that establish the technical direction for research infrastructure.
Requirements
- Build first versions of systems from a blank page with limited direction and write designs or specifications for others to implement.
- Have hands-on experience with large-scale distributed training, including multi-node GPU training and high-throughput data pipelines.
- Be fluent with container orchestration and ML lifecycle tooling such as Kubernetes, Docker, Prefect, and MLflow, or equivalent tools.
- Have engineered large-scale data systems for unstructured data, including storage, versioning, and preprocessing.
- Have taken a research prototype from a notebook to a system relied upon by others.
- Partner effectively with researchers through shifting priorities, failed experiments, and experimental iteration.
- Use and build AI tools to amplify engineering output and quickly create useful internal interfaces.
- Explain ML systems clearly to researchers and non-technical stakeholders.
- Preferred: hands-on training or evaluation of speech/audio models such as ASR, STT, or TTS, or LLM experience with a path to audio.
- Preferred: experience embedded with a research team or contributions to open-source speech or ML tooling such as ESPnet, NeMo, SpeechBrain, or WhisperX.
- Typically 5–12 years building software or ML systems, or equivalent demonstrated experience; an advanced degree or equivalent industry experience is preferred.
Tech Stack
Categories
Data EngineeringML Engineering
About Deepgram
Deepgram builds Voice AI infrastructure for developers, offering speech-to-text, text-to-speech, and voice agent tooling through real-time APIs and deployable on-prem software. Privately held and founded in 2015, the company is headquartered in San Francisco. Customers such as Twilio, Cloudflare, and Jack in the Box use its models to power production voice features; products include Nova-3 STT, Aura TTS, and a Voice Agent API.