Lightning AI debuts ‘iPhone approach’ to new AI dev platform

Lightning AI, the company behind popular open source framework PyTorch Lightning, has launched Lightning AI Studios, an enterprise-grade, all-in-one cloud-based platform for building and deploying AI products at scale. 

According to Lightning AI, the end-to-end, SaaS platform “fundamentally re-architects how developers work by abstracting away every non-core activity and providing a single interface for all AI development needs.” Users can use pre-built templates, quickly scale from CPU to GPU on as many machines as they need, and leverage natively integrated tools or build their own while deploying anywhere — in their cloud, Lightning AI’s cloud or on a local GPU cluster.

“If we’re successful. I hope that this is a moment like where you go from using your flip phone to an iPhone,” William Falcon, founder and CEO of Lightning AI, told VentureBeat during a virtual demo.

An operating system for AI developers

Falcon describes Lightning AI Studios as both a kind of operating system for AI developers, as well as a collaborative cloud-based desktop complete with all necessary apps in one place, similar to how a smartphone brings together a phone, a camera, a calculator and a variety of apps. According to Falcon, tooling for machine learning “has not meaningfully evolved in decades — until now.”

Three years ago, Falcon began work on Grid.ai, the precursor to Lightning AI, which raised $18.6 million in a Series A round to help AI researchers and engineers bring their models to production. At that time, the team introduced idea of training models in the cloud.

“My whole thesis for doing this was I just fundamentally believe that coding in your laptop is bad for AI,” he said. “We wanted to say, what’s next? What is the iPhone approach to this thing — we like to think of ourselves as an Apple for machine learning.”

An app marketplace for AI model production

Now, three years later, he believes that iPhone-like approach has arrived, in the form of an experience that is fully in the cloud. “Your laptop is really just connecting to a cloud machine,” he explained. Lightning AI Studios requires no setup, so AI developers can start building AI products with one click, on one screen — including rapid setup, easy integration, and a marketplace of apps for building, training, fine-tuning, and deploying models.

There are apps for self-service inference API for the leading open source LLMs (Llama2, CodeLlama, Mistral); fine-tuning LLMs on user data; full LLM pre-training on user data; Stable Diffusion (including recent variants like LCM); fine-tuning and generation; LLM evaluation; multimodal image + language chat API; and vision + audio + text quickstart.

Falcon said Lightning AI Studios is ready to compete with two main competitors, Google Colab and AWS Sagemaker. “Those now feel like ‘gen 1’ cloud products,” he said, like “how the Blackberry was fun until people used the iPhone.” He claimed there are already 27,000 people on the waitlist for Lightning AI Studios. “We recently started getting people off the waitlist, and we already have close to 1000 people that are brand new, that weren’t using the product before,” he said.

There are four pricing tiers: a free level for individual students, researchers and hobbyists; a pro level for engineers, researchers and scientists; a teams level for start-ups and teams; and finally an enterprise level for larger organizations in need of enterprise-grade AI.

To sign up for early access to Lightning AI Studios, go to https://lightning.ai.