Introducing Function

Introducing Function

Mass AI adoption requires a top-notch developer experience. Enter Function.

In a world where everyone is racing to leverage the power of Artificial Intelligence (AI) to improve their products and workflows, the biggest blocker to mass adoption isn’t in prediction latency, or model accuracy, or chat streaming — it’s developer experience.

We’ve been crafting a platform to enable extremely simple, rapid AI deployment; one that could scale up to millions of predictions, and down to zero when idle; one that you could use from your high- and low-code environments; and one that caters to a broad spectrum of users, ranging from large enterprise data teams to small businesses looking to leverage AI without any fuss.

Simplifying AI for All

Traditionally, deploying AI has required developers to have a deep understanding of complex technologies like Docker and Kubernetes. And while there have been attempts by other players to condense these down to custom packaging formats or custom YAML specifications, we’re convinced developers can get a much better developer experience. For many businesses, especially those without specialized engineering talent in these areas, the learning curve remains a significant barrier.

Creating a predictor on Function starts with simply writing a `predict` function.

In designing Function, we wanted a developer workflow that had exactly zero new concepts, and as little cognitive load as possible. What we settled on was only the second concept anyone new to coding would have to learn: functions. Simply bring a Jupyter notebook with a predict function somewhere in there, and we’ll spin it up to serve as many as millions of prediction requests.

Make Predictions from Anywhere

Creating a prediction function — or ‘predictor’ as we like to call it — is the easy part. The hard, but most rewarding, part is getting these predictors in front of as many people as possible. To that end, we’re launching with official clients for JavaScript (browser+Node.js), Python, and Unity Engine:

We currently provide clients for Python, JavaScript, and Unity. Swift and Kotlin coming soon!

But we’re pushing even further: we want to enable you to not only deploy AI to your customers, but to develop these AI features and products collaboratively with your team. To that end, we’re also launching with official Slack and Discord bots. With either one, simply use the /predict slash command to make a prediction right in any channel or thread conversation. No need to wait for a new CI build or Vercel deployment to share your AI features with your team; just have them /predict .

Generating an image with Stable Diffusion with Function for Slack.

Everyone can be an AI Developer

AI continues to advance at a fast pace, unlocking the potential to solve a broader range of tasks than ever before. Yet, for many, the mastery of complex coding remains a formidable barrier. Function is launching with the ability to generate prediction functions from a description of what you would like the function to do:

Using AutoFunction to create a prediction function that can search online.

This feature is open source, runs on Function, and is available to all users at no added cost.

Join the Beta

Function represents a leap towards democratizing AI to all: in expanding both what kinds of developers can leverage AI in their products; and who we call a developer in the first place. By stripping all the deployment chaos down to a single predict function, and providing a wide variety of ways to run aforementioned function, we hope that Function becomes the first thought for anyone looking to harness the power of AI.

Whether you are a seasoned data scientist or a small business owner with a keen interest in AI, you no longer have to wait for some fancy vertical-SaaS-AI startup to discover you and your problems. Start experimenting and deploying AI with your team and customers today, with just a single predict function. The future is simple with Function AI.

Register at