Research Scientist, Foundation Model

About the role

You will own the research question at the center of our foundation model of the physical world, focusing on how a model can learn site-specific patterns to identify anomalies as unexpected events. You will work alongside perception and model engineers to develop probabilistic site memories and world models that predict states in latent space. Your research results will be deployed directly into production systems running on live security sites.

Responsibilities

• Develop a learned, probabilistic model of each site to understand typical scene behavior.
• Replace rule-based alerting with measured surprise and develop evaluation methods for rare incidents.
• Post-train open world-model weights on multi-sensor field sequences using latent space prediction.
• Transfer site-specific memory to a generalized model across all sites and domains.
• Define research directions, design experiments, and validate findings before production deployment.

Required skills

• Machine Learning
• Computer Vision
• Robotics
• World models
• Self-supervised learning
• Probabilistic modelling
• Spatio-temporal data
• Anomaly detection
• PyTorch

Nice to have

• Scene graphs
• Multi-object tracking
• Trajectory forecasting
• Bayesian deep learning
• Embodied AI
• Simulation

Qualifications

• PhD in machine learning, computer vision, robotics or a related field
• First-author publications at top venues

Benefits

• Equity package
• Career growth opportunities

About the Company

Laelaps AI is a founding team based in Zurich building intelligent software for physical security. They aim to create an orchestration layer for autonomous systems to handle critical tasks in the real world.