Real-Time Deep Learning Predictions

trainML Endpoints give you the real-time predictions you need for production applications without having to worry about servers, certificates, networking, or web development.

Focus on Your Model, not Web Development

Machine learning projects aren't complete until there is model deployed doing production inference. In many cases, business requirements dictate the need for inference in the form of an always-on, sub-second response time REST API. However, this requires wrapping model code in a web application, deploying it in web server on a GPU-enabled instance, configuring the necessary security and networking, and all the other operational skills and processes associated with web hosting.

trainML Endpoints remove this barrier from your project. You simply provide us with your model code and artifacts, configure which functions should run when the endpoint is queried, and we handle the rest. Your model will run behind a production grade webserver with a dedicated GPU, and all communication is fully encrypted. With production deployment handled, you can get back to focusing on model design and performance.

Framework Agnostic

Others have tried to solve the model deployment problem by imposing constraints on how your model code is designed, what framework it uses, or the format it must be saved in. These tools end up costing countless hours to learn their proprietary syntax and refactor perfectly good model code just to get started. Those that opt for a "bring your own" method require you to package everything yourself and build a functioning web app in a container, which leaves most of the difficulty to you.

You can deploy a trainML endpoint with a model built in any framework or that uses any libraries. The endpoint seamlessly translates HTTP requests into function calls to any code you specify. There is no framework to adapt to, and no code changes to make. Requests and responses use industry-standard JSON, and can include anything allowable JSON datatype.