Privacy-First Meeting Transcription for Banks on OpenShift
Banks generate enormous volumes of critical conversational data every day — from client advisory calls and credit committees to trading desks and board meetings. For banks, transcription cannot be processed in external public clouds at all, due to regulatory, confidentiality, and jurisdictional restrictions. As a result, most native transcription in Google Meet, Microsoft Teams, and Zoom is unusable for regulated financial environments.
Vexa is a self-hosted, open-source (Apache 2.0) meeting transcription platform built specifically for banks and regulated financial institutions. It runs entirely inside your Red Hat OpenShift cluster, ensuring that audio, transcripts, and metadata never leave your controlled infrastructure.
This gives banks full data sovereignty, eliminates third-party cloud exposure, and removes vendor lock-in.
On-Premise by Design, AI-Ready by Default
Vexa performs all speech processing locally inside OpenShift and streams real-time transcripts directly into internal AI and analytics systems. It integrates natively with OpenShift AI and supports modern LLM stacks including Mistral, DeepSeek, IBM Granite (Queen), and Llama Stack.
This allows banks to deploy retrieval-augmented generation (RAG), compliance analytics, and conversational intelligence entirely on-premise, without exposing sensitive data to external AI providers.
Built for Regulated Environments
Vexa is currently being piloted by European financial institutions for on-premise regulatory use cases. The platform is progressing through Red Hat Partner Connect ISV certification with OpenShift compatibility validation underway. It is also used in production by large enterprise organizations within the Academy Software Foundation ecosystem, including Sony, Disney, ILM, and Autodesk.
This combination of financial and industrial adoption demonstrates both regulatory readiness and production-scale reliability.
How It Works
Vexa is deployed as containerized services inside your existing OpenShift cluster. Your systems request a transcription bot via API, the bot joins a Microsoft Teams or Google Meet session, processes speech locally on your GPU, and streams transcripts in real time to your internal AI pipelines. Zoom support follows the same on-premise model and is on the roadmap.
All processing remains fully inside your infrastructure.
Business Impact
At enterprise scale, public cloud speech-to-text services become cost-prohibitive. For approximately 10,000 transcription hours per month, cloud costs can exceed $28,000 monthly. A typical self-hosted Vexa deployment operates around $5,000 per month in infrastructure, delivering an estimated 83% cost reduction while simultaneously strengthening security and regulatory control.
Vexa is designed to support GDPR, MiFID II, and Basel III compliance by keeping all conversational data under direct bank control with full audit trails.
Technical Requirements
Vexa runs on Red Hat OpenShift 4.12+ and requires a CUDA-capable GPU with at least 4 GB VRAM. Typical deployment on an existing cluster takes one to two hours.
30-Day Banking Pilot
Banks can validate Vexa through a fully supported 30-day on-premise pilot, including assisted deployment, OpenShift AI integration examples, and compliance documentation.
Request access to the 30-day banking pilot.
Contact
Schedule a confidential on-premise demo:
https://cal.com/dmitrygrankin/30-min
Open source repository:
https://github.com/Vexa-ai/vexa
Dmitry Grankin
CEO & Founder, Vexa
https://www.linkedin.com/in/dmitry-grankin/