Gradient

Gradient is a developer of an application programming interface (API) platform designed for large language models (LLM) to allow companies to build custom and private artificial intelligence (AI) applications in the cloud. The company offers real-time data infrastructures and web APIs to allow developers to fine-tune and generate completions on open-source LLMs. This is intended to allow users in sensitive industries, such as healthcare and finance, to create AI solutions with private data that are compliant with SOC 2 and HIPAA regulations.

Gradient was founded in 2022 by Chris Chang, Mark Huang, and Forrest Moret after the trio worked on AI products at other technology companies and based on a belief that LLMs could be transformative for the enterprise if they had a reliable way to add private and proprietary data to them. To do so, Gradient developed a platform that hosts a number of open source LLMs, including Meta's Llama 2, which users can fine-tune to their needs and based on particular use cases, such as data reconciliation, context-gathering, and paperwork processing.

Gradient's platform is developed to allow users to fine-tune LLM models in the cloud with private and proprietary data. To do this, Gradient hosts the models in a single place, in Gradient's AI Cloud, to help reduce overhead while guaranteeing low latency and ensuring that models and data are secured and privacy is maintained, keeping them compliant with SOC 2 and HIPAA regulations, among others. This is intended to allow users in sensitive industries, including healthcare, finance, legal, and education sectors, to use LLMs without concerns.

Gradient also offers APIs to allow users to fine-tune their models through an API, including creating a private instance of a base model and instructing that model on private or proprietary data to see how the model learns in real time. These APIs are developed to be able to be embedded or integrated into other platforms as well, to allow users to have full control of the data used and what is generated from the platform, including ensuring no one outside of a private workplace can have access to those models and generations.