VAELLA VAELLA
All Jobs

AI Engineer— Perception & Spatial AI

About Vaella
Vaella is building the perception and decision layer for the buildings around us. Our LiDAR-based Vision Sensors give buildings the ability to see people anonymously and accurately, and our Building AI Operator turns that perception into action — controlling elevators, dispatching security, filing maintenance tickets, and drafting reports.

 

Three product surfaces work together: Vision Sensors (the eyes), the Digital Twin (the memory of the building), and Simulation (the foresight). Together they let operations teams run more complex buildings with the same headcount, with a 10× productivity target and a privacy posture that is architectural, not bolted on.

 

About The AI Vision Lab
The AI Vision Lab is the team that turns raw LiDAR depth into the structured signals everything else in Vaella runs on — occupancy, density, flow direction, dwell, queues, and anonymous tracks. We own the perception stack end to end: sensor design, on-device processing, the models that make sense of point clouds, and the spatial reasoning that grounds the Digital Twin and Simulation.

 

Privacy by physics is a hard constraint on everything we build. The sensors capture no images and no biometric data, and our models work from depth alone.

 

The Role
We are hiring an AI Engineer to work on perception models and spatial AI. You will design, train, and ship the models that let our sensors and platform reason about people in real-world buildings — from a single sensor stream up to a building-wide spatial picture.

 

This is a hands-on engineering role with a high product surface. You will see your work running in metro stations, airports, and offices within weeks of shipping it.

 

What You Will Work On

  • Perception models on point clouds. Detecting and tracking people from LiDAR data in dense crowds, with accuracy that holds up when bodies overlap and lighting shifts.

  • Multi-sensor spatial reasoning. Fusing data across many sensors into a single, coherent picture of a floor or a building, so the Digital Twin sees the whole space rather than a stack of zones.

  • On-device inference. Pushing models down to the sensor itself so raw distance measurements never leave the unit — working with the embedded constraints that go with it.

  • Behavioural signals. Going beyond counts to flow vectors, queue formation, dwell patterns, and anomaly detection that the AI Operator can act on.