New startup Continual Inc. secured a $4 million seed funding round on Thursday to support the rollout of its cloud service, which allows companies to build artificial intelligence models that improve over time.

In conjunction, the startup made the service available through a public beta program.

Continual’s $4 million funding round was led by Amplify Partners. Illuminate Ventures, Essence, Wayfinder and Data Community Fund participated as well. The startup will reportedly use the funding to double its headcount over the next year and plans to build new features for supporting common AI use cases.

AI models become more accurate over time by learning from the data they process. Facilitating this continuous improvement in accuracy can be complicated for enterprise software teams. Developers must regularly supply a neural network with new data that it can analyze to find ways of carrying out calculations more efficiently.

Continual promises to simplify the task. The startup’s service provides features that allow companies to refine their AI models using information they keep in cloud data platforms such as Amazon Redshift, Google BigQuery, Snowflake and Databricks.

Software teams can use Continual’s service to monitor their neural networks’ performance. When developers identify an opportunity to increase accuracy, they can use the service to retrain a machine learning model using new data. To save time for users, Continuous offers tools that make it possible to automate certain aspects of the AI training workflow. 

Continual’s service streams neural network calculation results back to the data platform from which it pulls the information used for AI development. Once the results are in the data platform, a company can easily sync them to business intelligence tools for further analysis.

“By writing directly back to your data warehouse, you can easily consume up-to-date predictions from existing BI, Reverse ETL, and downstream tools,” explained Continual co-founder and Chief Executive Officer Tristan Zajonc. “There is no new infrastructure or complex integrations to manage to make predictions actionable.”

Another task that Continual promises to help with is feature management. Features are the data points that an AI uses to make decisions. A neural network that estimates future customer demand, for example, might use data points about past purchases to generate predictions.

Continual makes it possible to define features using the SQL syntax, which is widely used in analytics projects, and removes the need for workers to use specialized AI development tools. Feature definitions can be organized in a centralized library to ease management.

Image: Continual