Data Science

New! Domino AutoML

Thomas Dinsmore2023-10-09 | 4 min read

Return to blog home

You've heard the pitch from AutoML vendors, and it sounds appealing. AutoML reduces delivery time and helps your team deliver more value. AutoML empowers users with business skills so your expert data scientists can spend more time on high-value work. With built-in quality control, AutoML helps your team deliver consistent and reliable results.

But you've tried most of the available products and found them wanting.

  • Standalone AutoML tools speed model training, but all your data science projects have complicated pipelines. You know what they say: data scientists spend 80% of their time wrangling data.
  • Your project teams include people with diverse skills. Even if AutoML can help with some tasks, that work is only helpful in the context of a larger project.
  • Most AutoML tools work with small datasets. Your team works with Big Data, and your models are complex.
  • Your team needs comprehensive reproducibility. Many AutoML tools are black boxes. That's not acceptable.
  • Some AutoML tools run on-prem, while others run in the cloud. That's great. However, you must train models on many platforms, cloud regions, and hybrid environments. That's a show-stopper for some tools.

With the Domino Summer Release, we engineered AutoML directly into the Domino platform. Domino AutoML provides you with the benefits of automation without the headaches:

  • Domino AutoML delivers editable code directly to your project workspace. You don't need to "shoehorn" executables from another application into your pipeline.
  • Domino projects can combine custom and AutoML models so your expert data scientists can collaborate with diverse users.
  • Domino AutoML can leverage Domino ephemeral Spark and Ray clusters for scale-out processing. There are no pre-set limits on the size and complexity of the models you can build.
  • Domino AutoML works seamlessly with Domino's industry-leading reproducibility engine. Domino automatically captures every experiment and every model. That means your team saves time by leveraging previous work. And you can reproduce a project precisely for regulatory compliance.
  • Domino AutoML runs on Domino Nexus, our unique capability for hybrid and multi-cloud workloads. Your team can run AutoML experiments in any cloud, region, and computing platform.

Many existing AutoML tools use a proprietary algorithm to find the best model. Domino AutoML uses FLAML, an open-source library from Microsoft Research. FLAML (Fast and Lightweight AutoML) is a Python library that automates model selection and hyperparameter tuning. It uses a technique called Cost-Frugal Optimization to optimize model quality at the lowest possible cost.

Domino AutoML settings configuration

We built FLAML into Domino Code Assist (DCA). DCA provides an intuitive point-and-click interface for common analysis and data science tasks. Domino AutoML generates code to prepare the data, test algorithms, and optimize hyperparameters. Your users can modify the code to improve the model if needed. And your team can embed the code in a Domino workflow for deployment.

Domino Sentry governs Domino AutoML. Domino automatically captures all of your team’s work for guaranteed reproducibility. With Domino's approvals workflow, you know that every data science product gets the required level of review and approval.

Use Domino AutoML for machine learning that is fast and governed.

Learn more about Domino AutoML here.

Thomas Dinsmore is Domino's Head of Competitor Intelligence. He is passionate about driving product vision and enabling the field.

Subscribe to the Domino Newsletter

Receive data science tips and tutorials from leading Data Science leaders, right to your inbox.


By submitting this form you agree to receive communications from Domino related to products and services in accordance with Domino's privacy policy and may opt-out at anytime.