Hey HN, Adam here, one of the founders of Booklet (https://booklet.ai).<p>As a previous machine learning team leader, we would have a team celebration every time an API endpoint was created to serve a new model, but a model behind an API endpoint was rarely useful for other teams. Marketing wanted LTV predictions in their sales tool, the email team wanted customer segments in their email tool, and operations wanted sales forecasts in their warehouse management system. Those integration steps were painful and slow for the data team to build, especially with proper production-ready stability.<p>We built Booklet.ai to make these post-deployment integration steps easier. We connect to existing ML endpoints (on Sagemaker or any other tool) and you can set up a quick testing UI for spot-checking. From there, set up sources (ex: AWS Redshift) to feed into the model for predictions and destinations (ex: Intercom) for those predictions to be sent to. These come together into a dataflow that is fully managed - scheduling, error handling, retries, etc.<p>If you want to test it, try our public demo, which is applied to a lead scoring model sending results to Redshift and Intercom (https://app.booklet.ai/model/lead-scoring). To follow along the tutorial of how we built this model, check out our blog post here (https://towardsdatascience.com/a-true-end-to-end-ml-example-lead-scoring-f5b52e9a3c80).<p>We’d love to hear your feedback! Please let us know if you have questions, thoughts or feature ideas. Thank you!