MLOps Engineer: Reproducible AI Pipelines

Harmattan AI; Zürich; Vor Ort; CHF 90.000 - 150.000; Vollzeit

Zusammenfassung

Harmattan AI seeks an MLOps Engineer (Zurich/Lausanne/Paris) to own the training and evaluation pipelines, CI, and reproducibility. You will enable modelers to focus on models by standardizing infrastructure and tooling.

The role emphasizes responsible, scalable ML in a defense context, with cross-location collaboration and production-grade ML workflows. The candidate will contribute to a secure, auditable pipeline environment, enhance experiment tracking, and promote reusable templates while

Qualifikationen

• Degree in a STEM field or equivalent practical experience.
• Experience building ML pipelines, CI, and experiment tracking.
• Proficiency in Python and infrastructure software.
• Bonus: experience with on-device or hardware-in-the-loop testing.
• Attributes: systematic, reliability-minded, and service-oriented.
• Commitment to Harmattan AI's mission of defensive, high-impact tech.

Aufgaben

• Build and maintain reproducible training and evaluation pipelines for models.
• Introduce CI to ML work and ensure runs are comparable across code, config, and models.
• Maintain a model lifecycle registry from sandbox to production.
• Keep results traceable and comparable through experiment tracking and logging.
• Wire on-device and hardware-in-the-loop testing into automation and collect results.

Kenntnisse

Python

CI/CD

ML pipelines

Experiment tracking

Software engineering

Ausbildung

STEM degree

Tools

Model Registry

Jobbeschreibung

Harmattan AI seeks an MLOps Engineer (Zurich/Lausanne/Paris) to own the training and evaluation pipelines, CI, and reproducibility. You will enable modelers to focus on models by standardizing infrastructure and tooling.

The role emphasizes responsible, scalable ML in a defense context, with cross-location collaboration and production-grade ML workflows. The candidate will contribute to a secure, auditable pipeline environment, enhance experiment tracking, and promote reusable templates while