PtSe₂ is a promising 2D material for nanoelectromechanical sensing and photodetection in the infrared regime. One of its most compelling features is the facile synthesis at temperatures below 500 °C, which is compatible with current back-end-of-line semiconductor processing. However, this process generates polycrystalline thin films with nanoflake-like domains of 5–100 nm size. To investigate the lateral quantum confinement effect in this size regime, a deep neural network is trained to obtain an interatomic potential at density functional theory accuracy and it is used to model ribbons, surfaces, nanoflakes, and nanoplatelets of PtSe₂ with lateral widths between 5 and 15 nm. Which edge terminations are the most stable are determined and evidence is found that the electrical conductivity is localized on the edges for lateral sizes below 10 nm. This suggests that the transport channels in thin films of PtSe₂ might be dominated by networks of edges, instead of transport through the layers themselves.