Backprop

Congrats on the launch! I'm really excited to see how this goes. I'd love to see a json + CLI (like for PaaS configs) for setting up CI/CD and delivering models in the future. Good luck!

Hi, Product Hunt! I'm Owen, a founder and developer at Backprop. When integrating machine learning into software, deployment is a hefty investment — in terms of both time and money. Our goal is to build a platform that makes using state-of-the-art models in production easy, scalable, and affordable for every developer. We've made deploying models as painless as possible: once you've trained one, you just prepare a .zip file, and drag it onto our dashboard's upload page. In just a few minutes, your model will be accessible to you via an API. That's it — just a few steps, and you're ready to go. Plus, if you make changes and want to update a hosted model, we offer CI/CD: your old model will be available with zero downtime until the new version has been built and deployed. On top of that, we have several pre-trained models available on our API for tasks out of the box - OpenAI's GPT-2, OpenAI's CLIP, Google's T5, and Facebook's BART, just to name a few. We're constantly on the lookout for new research so we can expand our selection of included models & tasks. We operate on a serverless architecture: rather than having to pay for infrastructure that sits idling, Backprop charges you for the time computation time your models take to solve tasks. That way, you're only paying for the time where you're deriving value from our API. Since we also want to contribute to the ML developer community, we've also built an open-source Python library that makes locally using and finetuning these state-of-the-art models a simple process. Once you have your data, you can train models in a single line of code, with the heavy lifting done by us under the hood. You can upload models to the platform via the library in a single line, too — but the library works independently from the platform just as well. See it here (and leave us a star!): https://github.com/backprop-ai/b... Thanks for reading! Give us a shot, and feel free to reach out with any feedback and thoughts. We'd love to hear what you have to say. If you shoot me an email after you've signed up, I'd be happy to extend your trial period so you have more time to experience what we have to offer. Owen LaCava owen@backprop.co

Hey everyone! I'm one of the makers of Backprop and wanted to give some ideas on what a machine learning API such as ours can be used for. We feature many tasks, such as text generation, image vectorisation, and question answering. If you don't work with AI on a daily basis, then these might not mean much to you - so here's some use cases for some of our supported tasks: Text generation - this endpoint is like GPT-3. While less powerful, it's available for everyone to use. Solve a large variety of tasks from generating ideas and poetry to responses for customer messages. Image vectorisation - make an image search engine that works via text or other images. Question answering - chatbots for customer service/sales. These are just some examples and there are many ways to get creative. If you got ideas for using AI in your app, but don't exactly know where to begin then we'd love to help you out! We're constantly on the lookout for new exciting use cases, so don't hesitate getting in touch. PS, if you have any tasks you'd like us to support, then let us know.