Locai

Locai is a private AI infrastructure platform that lets organisations run modern AI models entirely on hardware they own and manage. Rather than using cloud inference, it executes language, vision, audio and multimodal models on local servers, private clouds, edge devices or fully air-gapped systems, keeping data and computation inside your environment.

The platform is organised into three parts: Workspace, the user interface for tasks like document generation, database queries, transcription and retrieval; Control, which handles deployment and orchestration across your infrastructure; and Link, which provides API access so applications can call locally hosted models. Together they form a single environment for deploying and operating on-prem AI without routing inference to external providers.

Locai is aimed at organisations with strict security or regulatory needs. It supports completely offline or air-gapped deployments so sensitive documents, meeting recordings and operational data can remain on-premise. The platform supports multiple model types — large language models, speech recognition, computer vision and video processing — so outputs from one model can feed another under the same management layer.

• Supports model formats such as GGUF, ONNX and TensorFlow Lite
• Can deploy from an integrated registry or import custom models
• Runs on NVIDIA and AMD GPUs, Apple Silicon and CPU-only systems

More information at https://www.locai.co.uk/