ML Ops Engineer – Real-Time Voice AI Infra

Jobicy; United States; A distancia; USD 140.000 - 195.000; Jornada completa

Descripción de la vacante

Deepgram is seeking an ML Ops Infrastructure Engineer to bridge research and production. You will design pipelines, deployment systems, and testing infrastructure to scale models for real-time voice APIs.

Collaborate with research and engineering to ensure reliable, low-latency inference across environments. You’ll build automated retraining, observability dashboards, and canary deployment strategies to improve model health and cost efficiency.

Formación

• 4+ years of experience in MLOps, DevOps, or infrastructure engineering focusing on ML systems.
• Strong Python skills and automation for ML workflows.
• Proficiency with CI/CD and pipelines for software and model delivery.
• Hands-on experience with Docker and Kubernetes for containerized workloads.
• Experience deploying ML models in production and monitoring them.

Responsabilidades

• Design and build CI/CD pipelines tailored for ML model development, validation, and deployment.
• Architect and maintain model deployment pipelines from research to production.
• Build A/B testing infrastructure for controlled rollouts and measuring impact.
• Implement monitoring for model performance in production (accuracy, latency, drift).
• Develop automated retraining pipelines triggered by data changes or performance drops.
• Create and maintain environments mirroring production for researchers.

Conocimientos

Python

CI/CD

Docker

Kubernetes

ML Ops

Herramientas

Terraform

Pulumi

Prometheus

Grafana

Datadog

ONNX Runtime

NVIDIA Triton

Descripción del empleo

Deepgram is seeking an ML Ops Infrastructure Engineer to bridge research and production. You will design pipelines, deployment systems, and testing infrastructure to scale models for real-time voice APIs.

Collaborate with research and engineering to ensure reliable, low-latency inference across environments. You’ll build automated retraining, observability dashboards, and canary deployment strategies to improve model health and cost efficiency.