LogoLogo
  • About
  • Privacy Statement
  • Sustainability
  • Terms of Use
  • Getting Help
  • Info for Pilot Participants
  • Services
    • Transcribe
      • Speech-to-text Models
    • Ask Oscar
    • LibreChat
    • Upcoming services
Powered by GitBook
On this page
  • Overview
  • Motivation
  • Architecture

Was this helpful?

Export as PDF

About

NextPrivacy Statement

Last updated 1 month ago

Was this helpful?

Overview

CCV AI Services (), informally known as ChatCCV, provides a set of generative AI tools that are built and deployed in-house at Brown for all members of the Brown community. These tools include:

  • (currently under construction): A unified chat interface that provides the community free and private access to a mix of state-of-the-art open-source and commercial large language models (LLMs), including LlaMA 3.3, ChatGPT-4o, Gemini 2.0 Flash, and DeepSeek V3.

  • : A web app that provides fast and accurate audio transcription through open-source and enterprise-level models, including the OpenAI Whisper model and the Azure AI Speech-to-Text model.

  • (currently under construction): An AI assistant that can answer questions based on documentation of CCV and other Brown services.

Motivation

The main motivation for providing in-house generative AI tools at Brown is data privacy and security. Commercially available AI tools such as OpenAI's ChatGPT and Anthropic's Claude are not fully transparent about what they do with the data that the users provide them, including chat history, documents, images, and videos. The data may be used by these commercial companies to train their next model, risking exposure of sensitive information. With AI services provided by CCV, we have full control over how user data is handled and can be completely transparent about how and where we store, collect, and use user-provided data. Please see our section for details about how user data is stored.

Another motivation is to provide AI services tailored specifically to the needs of the Brown community. Our service, for example, was developed based on popular demand from the Brown community for a high-performance, reliable, cost-effective, and secure audio transcription service. We understand that one of the major concerns of the community about generative AI tools is sustainability, so we built a carbon meter for some of our services to estimate the carbon footprint of AI usage. We will continue to listen to the community for their feedback on how to improve our services and develop new services that meet their need for an ever-evolving landscape at work.

Architecture

The CCV AI Services were built with data security and privacy in mind. To that end, the service are deployed on infrastructure owned and/or managed by Brown. Specifically, the LIbreChat service is deployed on Brown's own Kubernetes cluster with connection to AI models provided by Brown's cloud partners Google Cloud Platform (GCP) and Microsoft Azure. The Transcribe and Ask Oscar services are deployed on GCP with additional APIs and services provided by Microsoft Azure. Below are the architecture diagrams:

https://ai.ccv.brown.edu
LibreChat
Transcribe
Ask Oscar
Privacy Statement
transcribe
Architecture diagram for Transcribe and Ask Oscar
Architecture diagram for Transcribe and Ask Oscar