Salesforce hopes to make it easier for companies to train AI models. The initiative is called Bring Your Own Model. Einstein Studio allows customers to leverage their data sets on Salesforce Data Cloud to run their own models on them.
Salesforce is first supporting models built with Amazon SageMaker. This allows companies with such models to deploy their solutions without a large internal hardware infrastructure. In the future, Salesforce hopes to offer support for Google Vertex AI and more.
Further evolution
Salesforce talks about direct access to Salesforce data without having to copy it. Customers must use a Python connector to connect this data in an Amazon SageMaker notebook. According to the company, a user-friendly interface exists to run queries against the Data Cloud and get data in Python form.
The tool also works with LLMs to, for example, generate an automatic email when a product needs to be serviced before it breaks down, reports TechCrunch. Salesforce could curb hallucinations of such models by connecting the model to a graph database created with Salesforce data. An intermediate step like this is used more often to make LLMs more reliable, although the technology is difficult to get completely in line.
Also read: Salesforce and Google Cloud integrate even more AI features