Eye Tracking is based of bio-metric measurements, and the accuracy of the measurement will vary based of the individual. That means precautions have to be taken to ensure a large audience can utilize the technology. Enter selection algorithm. Given an xy coordinate or xyz direction from the eye tracker and some knowledge about what objects are gaze-aware you can start to try and pick the right object. This may seem trivial initial, however years of experience working with world class eye tracking devices, this is still a complex task.

Gaze to Object Mapping (G2OM), is such selection algorithm that aims to widen audience that can reliably use eye tracking. The core is centered around a machine learned algorithm implemented in the programming language Rust. I worked on the initial proof of concept and first production ready version with algorithms, training, annotation and infrastructure. End result was the best measurable selection algorithm in the company.