Foundational ML Engineer for Embodied AI in Security

Laelaps AI; Zürich; Vor Ort; CHF 140.000 - 210.000; Vollzeit

Zusammenfassung

Laelaps AI in Zürich is building autonomous security systems powered by foundation models and multimodal perception. As a Foundational Models & ML Engineer, you will train, fine-tune, and deploy VLAs, VLMs, and world models on real robots in real environments, balancing data, latency, and safety constraints.

You will collaborate with autonomy and platform teams to close the loop from data collection to model training to fielded deployment, shipping production-ready models that operate under

Qualifikationen

• PhD or equivalent research experience in ML, CV, robotics or related field.
• Strong PyTorch or JAX skills; comfortable with multi-GPU / multi-node scaling.
• Experience shipping foundation model work to production in embodied/multi-modal settings.
• Publications at top ML or robotics venues.

Aufgaben

• Train and fine-tune foundation models for robot control, perception, and reasoning.
• Build data pipelines from field data into training datasets.
• Run experiments at the intersection of foundation models and embodied AI.
• Optimize models for edge deployment: latency, quantization, distillation.
• Define interfaces between learned components and autonomy modules.
• Drive continuous evaluation from sim benchmarks to field metrics.

Kenntnisse

PyTorch

JAX

Robotics/Embodied AI

Model deployment

Data-driven evaluation

Ausbildung

PhD or equivalent research experience

Tools

IsaacSim

MuJoCo

Gazebo

Git

Jobbeschreibung

Laelaps AI in Zürich is building autonomous security systems powered by foundation models and multimodal perception. As a Foundational Models & ML Engineer, you will train, fine-tune, and deploy VLAs, VLMs, and world models on real robots in real environments, balancing data, latency, and safety constraints.

You will collaborate with autonomy and platform teams to close the loop from data collection to model training to fielded deployment, shipping production-ready models that operate under