The potential of machine learning is overwhelming. Predictive models promise to give insight into our world and business at levels of accuracy, scale and performance that have previously been unthinkable.
The potential is great, however each application requires serious consideration as there are no silver bullets (or free lunches!)
We work closely with you to understand your problem domain and to consider how best (or even if) to deploy a machine learning solution.
The biggest models are able to capture more of the underlying detail and subtlety across the problem space. Now that we have the tools to train neural networks using hardware such as GPUs we can create bigger and better machine learning models.
However such power is not always required. We like to create a baseline with simpler models such as Naive Bayes or K Nearest Neighbors that can capture the key details in your domain with faster training and inference time.
Even without labelled training data it's still possible to use machine learning in such areas as dimensionality reduction for visualization and clustering to automatically find patterns in data.