Enterprises face challenges in deploying large language models (LLMs) due to high computational costs, latency issues, and concerns about data privacy and security. Fine-tuning and deploying these models often require significant infrastructure investments and specialized expertise.

Proxis offers a platform for LLM distillation, fine-tuning, and optimization, enabling enterprises to deploy customized models on-premises or in the cloud with improved efficiency and security. The platform utilizes kernel-level enhancements and Triton integration to optimize model performance, resulting in faster inference speeds and reduced computational costs. Proxis also provides tools for automated AI agent creation, specifically for managing and responding to emails, by ingesting email history and company knowledge bases. The platform supports both hosted and on-premises deployments, ensuring data confidentiality and control.

LLM distillation and fine-tuning platform for customized model creation

Kernel-level optimizers that rewrite kernels into Triton for accelerated compilation times

Automated AI agent creation for email management, including drafting and sending capabilities

Integration with company knowledge bases for context retrieval

On-premises deployment options to ensure data privacy and security

VSCode extension for optimization function calling

Support for multiple models

The primary target audience includes enterprises and developers looking to deploy and optimize LLMs for various applications, with a focus on those requiring on-premises deployment for data security, as well as sales teams looking to automate email communication.

Proxis offers a tiered pricing model, including a free open-source option for developers and startups, and a Pro version for enterprises with enhanced AI capabilities, dedicated support, and custom agent optimizations.