Training an ML model requires experimentation and patience. So too does building a platform for creating ML-powered applications. We’re excited to share some of this month’s biggest wins toward empowering data scientists to build and maintain full-stack applications that deliver real value.
If you have a wishlist for Baseten features, please let us know! We hang out at email@example.com.
What we’re reading and writing
- A twitter thread by Santiago on deploying machine learning models
- Four models to accelerate content creation
- An open-source 100-billion parameter text generation model
- How to embed machine learning demos in your landing page
View builder level-up
Baseten gives data scientists full-stack superpowers. And using a superpower should be fun! It should make you feel, well, powerful. A critical mass of recent UX upgrades to the view builder means that building a webpage with Baseten now feels like flying.
Three new keyboard shortcuts carry the view builder experience: delete, undo/redo, and copy/paste. Each of these features works exactly as you would expect, and together they make experimentation easy and low-risk. Plus, smaller changes in the grid system and drag-and-drop logic mean that view building is a smooth, fast, predictable process.
The state of state
One of the challenges of building a full-stack application is everything you have to keep track of. React, one of the most popular web development frameworks, relies on Redux to manage application state. Meanwhile, dozens of logging solutions compete to let you know what’s going on in the backend.
While Baseten abstracts away most of that complexity, both application state and logs hold information for developing and debugging your application. Baseten now shows badges next to both logs and the data explorer (which tracks various types of state) to indicate the presence of new information that may merit your attention as you develop.
A warm welcome
We’ve onboarded hundreds of users since launch, and we want every minute spent using Baseten to be creative, not confused. To flatten the learning curve, we’ve added more in-product cues to accompany our ever-expanding documentation.
We’ve revamped the dismissable info cards shown to new users throughout the application builder to provide a more comprehensive tour. But what about when you need a refresher on specific details when building your second or seventy-second app? We’re adding subtle hints throughout the product, starting with tooltips on many component properties. These quick reminders link to specific documentation for a comprehensive reference.
This month, we were thrilled to be joined by Bola, our first forward-deployed engineer. We might be looking for you, too! We’re hiring for a developer advocate, lead infrastructure engineer, and lead ML engineer.
There is plenty of exciting stuff coming up in July and beyond, we can’t wait to tell you all about it next time!
The team at Baseten
Choosing the right horizontal scaling setup for high-traffic models
Horizontal scaling via replicas with load balancing is an important technique for handling high traffic to an ML model. Let’s examine three tips for understanding how to properly replicate your instances to save users time without wasting your money.