How Gradio Turbocharges your Machine Learning Team
Business use cases for Gradio
Gradio is being used by some of the world’s leading companies and machine learning teams, from Stanford’s Human-Centered AI to Cisco AI. How do they use it, and how can Gradio accelerate your machine learning (ML) team’s workflow? In this tutorial, I give a brief introduction to Gradio and outline 3 business use cases where Gradio allows you to do weeks of engineering work in under an hour.
Intro to Gradio
Gradio is an open-source Python package that automatically creates fully-functioning web apps with GUIs around ML models, in just 3 lines of code. With Gradio’s interfaces, non-ML users can make predictions using models without having to write any code right from their browser. This lets anyone give feedback and test ML models anywhere!
For example, try out a state-of-the-art pose estimation model with Gradio here: https://www.gradio.app/hub/hub-openpose
So how can businesses make the most of Gradio’s tools? Here are three use cases that we see.
Use Case 1: Immediate feedback from domain experts
As machine learning developers, we often create models to make predictions in specialized domains, such as medicine or finance. As we build models, we seek feedback from domain experts to ensure that the models are making sensible predictions on the full distribution of data.
At a leading public biotech company, the senior machine learning engineer recognized the potential that Gradio provides in accelerating this kind of feedback. Before using Gradio, the machine learning engineers and chemists would have regular lengthy Zoom calls to discuss which molecules the model was correctly predicting as toxic, and which molecules were being incorrectly flagged.
Rather than scheduling weekly video conferencing, and gathering iterative feedback over the course of several months, the company realized that Gradio would allow the ML developer to easily create a GUI that the chemists on the team could use to interact with the model and instantaneously flag anomalous predictions.
(By the way, this use case was the original motivation behind Gradio, as a tool for streamlining academic collaborations between ML researchers and radiologists. This has led to several peer-reviewed publications featuring Gradio.)
Use case 2: Streamline engineer-product communication
The design of a machine learning system, particularly which features to use as inputs and outputs, is not just a decision taken by the engineers on the ML team. Usually, the product team will provide feedback based on customer interviews. To provide this feedback, the product team will need access to the model in a way that they can use it without having to write code.
ML engineers will often spend lots of time creating demos and hacking together systems to gather feedback. Instead of reinventing the wheel, use Gradio’s ready-to-use UIs.
As Areeba Abid, a software engineer at Google’s Android SDK team, noted that Gradio can save teams that evaluate machine learning models in conjunction with product managers tremendous amounts of time. Similarly, Mohamed El-Geish, director of artificial intelligence at Cisco, is using Gradio across his teams to save them effort in creating demos for product decisions. As he put it:
“Gradio is now an essential part of our ML demos. All what it takes is a few minutes to make a demo come to life.”
— Mohamed El-Geish, Director of AI at Cisco
Use case 3: Create customer & sales demos
Gradio is just as useful for external demos as it is internal ones. Whether you’re a consulting company making demos for corporate clients, or an applied machine learning company creating an app for end customers, you can help close the sale by creating GUIs that let your customers immediately see value in your model. Rather than spending many months hiring a front-end engineer and a deployment engineer, Gradio lets your ML developer create this demo in under a minute.
You can also create GUIs that compare models, like in this image recognition model above. You can use this to showcase the key differences between your models, or demonstrate how your model is better than the competition, and close that sale!
If you’re interested in trying out Gradio, just visit https://github.com/gradio-app/gradio to see the open-source library. Or reach out to us at team@gradio.app.
