Private AI for financial services
From GPUs and infrastructure to private inference, secure workspaces and AI engineers, PrivateMind gives financial services firms everything they need to run AI privately. It runs globally, and your prompts and outputs are never used for training.
Anyone can sign up and try PrivateMind under our fair use policy.
Key facts
- Private by defaultPrompts and outputs never used for training
- Sovereign by designYou choose where data is held
- Bare metal NVIDIA GPUsRTX PRO 6000 to B300
- Open-weight by defaultFrontier models only if you enable them
- Operated around the clock24/7 by Options Technology
The stack
Five products in one private stack
PrivateMind is a neocloud built for financial services. We take the data centre, the GPU infrastructure and the operations off your hands. Use one part of the stack or all of it, and your data stays inside the same private environment throughout.
GPU & Infrastructure
Bare metal access to NVIDIA B300, B200, H200, H100 and RTX PRO 6000 GPUs in dedicated AI data centre space, with managed networking, storage and 24/7 operations.
- From part of a node to whole pods of GPUs
- Up to 800G Ethernet, isolated by default
- Block, object and high-performance storage with VAST and Everpure
Control Plane
Everything your ML engineers need to train, register and serve models on dedicated GPUs we run for you, without having to become infrastructure specialists first.
- GPU allocation down to MIG slices, with drivers built in
- Training with Ray on Kubernetes, tracked in MLflow
- Custom inference endpoints for the models you train and host
Workspaces
Agentic coding environments on GPUs and CPUs reserved for your firm, with VS Code, a terminal, OpenCode and Jupyter ready to go. Use them in the browser or from your own machine through the SDK and CLI.
- Coding agents run in a controlled environment on our GPUs
- PrivateMind API keys are minted automatically, so setup takes minutes
- Quant templates with market data from AtlasFeed already connected
Platform
Private LLM inference through chat and an OpenAI-compatible API, with cost-effective open-weight models alongside Claude, OpenAI and Amazon Bedrock, and every client kept in its own database.
- Hybrid routing across open-weight and frontier models, in chat and the API
- Model aliases that route around an unavailable model
- Separate databases by client
AI Solutions
Our AI engineers work with your people to find the workflows where AI saves the most time or removes the most risk, build the solution with them and keep improving it.
- Deployed AI engineers embedded in your teams
- Workshops for leadership, analysts and engineers
- A clear path from first use case to production: discover, deploy, advance
Playground
Try the models
Send up to three prompts free, no account needed. Then sign up for an API key.
Image, speech, OCR and embedding models are available through the API. Full catalogue
Kimi K3 2.8T-104B
What are we figuring out?
Own your data. Own your intelligence.
Create an API key after you sign up, then set it as PMIND_KEY. Any OpenAI-compatible SDK works the same way.
Prefer to start in code? Read the API quickstart or the Python SDK docs, or see what the Platform offers developers.
How we compare
The privacy of building it yourself, without the build
Public AI services are quick to start but hard to approve. Building private AI in-house takes months and a team to run it. PrivateMind gives you the control of an in-house build, ready to use from the first day.
| Capability |  | Public AI APIs | Building it yourself |
|---|---|---|---|
| Your prompts and outputs are never used for training | Never, on any plan | Depends on the plan and settings | Yes, if you build it that way |
| Open-weight models on hardware outside the public cloud | Our GPUs, in the regions you choose | Runs in the provider’s cloud | On hardware you buy and host |
| Frontier models when a task needs them | Hybrid routing, only if you enable it | The provider’s own models | Separate contracts and keys |
| Client data separated by database | Separate databases by client | Usually shared infrastructure | If you design and run it |
| Bare metal NVIDIA GPUs | Fractional nodes to multiple pods | Not usually available | After procurement and build |
| Networking, security and 24/7 operations | Managed for you | Only for their own service | Your team runs it |
| AI engineers working inside your team | Deployed AI engineers | Not offered | You hire them |
| Cost as usage grows | Move to dedicated GPUs and pay less per token | Per token, however much you use | High upfront cost before any value |
| Time to get started | Minutes on pooled capacity | Minutes | Months |
Built for a financial services security review. Read about security
Questions
Common questions about private AI
Anything else? Talk to our team or read the documentation.
What is PrivateMind?
PrivateMind is private AI for financial services. It combines bare metal NVIDIA GPU infrastructure, an MLOps control plane, secure agentic workspaces, a private LLM inference API and chat, and deployed AI engineers, all run by us and powered by Options Technology.
What is a neocloud, and how is PrivateMind different from a hyperscaler?
A neocloud is a cloud provider built specifically for AI workloads rather than general computing. PrivateMind is a neocloud designed for financial firms: bare metal GPUs, networks isolated by default, separate databases by client and no use of your data for training.
Can we run private LLMs without sending data to a public cloud?
Yes. Open-weight models run on PrivateMind infrastructure and are called through an OpenAI-compatible API or our chat app. When a task needs a frontier model, hybrid routing to Claude, OpenAI or Amazon Bedrock is available and only used when your firm chooses it.
Can PrivateMind meet data sovereignty requirements?
You choose where your workloads and data live, which helps you meet UK and EU GDPR and outsourcing requirements on where data is held. Our team can take your security and compliance reviewers through the detail.
Built with our partners
Start free today, or request a demo of private AI for your firm
Create an account to use chat and the API straight away, or request a demo to see dedicated capacity, workspaces and engineers for your team.