At Pencil, we are primarily a model aggregator, meaning we do not train our own foundational AI models. Instead, we use pre-trained models provided by third-party sources. This helps minimise the environmental impact associated with large-scale model training.
Training Our Own ML Models
In instances where we do train our own machine learning models (such as for ad performance prediction), the environmental impact is minimal. We take active steps to reduce this impact by:
Optimising cloud utilisation – This ensures training processes are as efficient as possible, reducing unnecessary resource usage.
Maximising training efficiency – We focus on improving the performance of the models with minimal computational resources.