Lightning AI

Lightning AI offers an LLMOps platform for customers to pre-train, fine-tune, and deploy LLMs in preferred cloud environments and host AI apps on private or on-premises infrastructure.Notably Lightning AI is the creator of PyTorch Lightning, a framework for training and finetuning AI models.

The company offers tools to streamline AI development, including dataset creation, parallelized tasks, optimized training, and remote dataset access. Key features include fault-tolerant streaming, persistent cloud-based coding environments, and seamless GPU integration. 

In 2023, the company launched LitGPT, a suite of 20+ high-performance LLMs with tools for pre-training, fine-tuning, and deployment. In 2024, Lightning AI introduced Thunder, a compiler to accelerate PyTorch models, and LitServe, a high-speed AI model serving engine. Lightning started as a private research library for generative adversarial networks (GANs) and variational autoencoders (VAEs), initially using Theano (2015) and later, TensorFlow. In 2017, it was restructured for PyTorch and renamed to “PyTorch Lightning.”

Key customers and partnerships

As of November 2024, the company had over 200,000 users of its platform. Notable clients included Google, Cisco, and Lexset. 

In February 2024, the company partnered with AWS to leverage its compute services and power GenAI services. In September 2022, the company partnered with NVIDIA to develop Thunder, a source-to-source compiler designed to optimize training and serving of GenAI models across multiple GPUs.