AI that investigates, not just flags, and never leaves the building.
Ratioc reads communications, reasons through each case against your policies, and produces signed evidence. It's built for compliance and surveillance teams who can't send their data to a third-party model.
Every other AI compliance tool asks for your data first.
Trading floor chat, voice transcripts, internal email. This is the most sensitive data your firm holds, and most surveillance AI wants it sent to a shared cloud model before you can even run an evaluation. For a fund operating under market abuse rules, conduct regimes, or LP-mandated data residency clauses, that question goes well past procurement. It reaches the board, and it usually ends the conversation before it starts.
We built Ratioc the other way round. The model, the data, and the evidence all stay inside your perimeter, from the first message ingested to the final sealed pack.
Three steps, one perimeter.
Identities are protected first
Every message is pseudonymised the moment it arrives. Analysts work with anonymised conversations; only an approved reveal workflow can re-identify a person, and every reveal is logged.
Reasoning grounded in your policies
A self-hosted language model checks each flagged conversation against your actual policy library. What comes back is a structured decision — the allegation, the reasoning, the policy it relied on, and a confidence score. Not a wall of prose.
Sealed and ready for audit
When a case is closed, its findings, signals, and policy references are frozen into a signed evidence pack. It's ready for internal review, an audit, or a regulator request.
Firms where the data can't leave.
Hedge funds
Communications surveillance across chat, email, and voice, without sending trader conversations to a third party.
Sovereign Wealth Funds
AI-assisted compliance reasoning that meets sovereignty and national-security data requirements without carve-outs or exceptions.
Investment banks
An additional layer of AI-assisted review on top of existing surveillance tooling, deployed inside your own data centre.
Regulated fintechs
Audit-ready evidence for conduct and market-abuse obligations, sized for a compliance team without a dedicated surveillance desk.
Your hardware. Your model. Your evidence.
Ratioc runs on dedicated infrastructure inside your environment. There's no shared cloud tenant and no multi-tenant SaaS instance. The reasoning model is self-hosted, so nothing depends on a third-party AI provider and there's no path for internal communications to leave your network.
The platform currently runs on dedicated on-prem infrastructure with over 50 days of continuous uptime across the full stack.
Built by engineers. Shaped with practitioners.
Ratioc is built by a small engineering team in London with 8 years of combined experience building data and AI systems, including in regulated financial services. What we care about is the unglamorous part of compliance technology: where the data lives, who can see it, and whether the person making the final call can see why.
Compliance and surveillance practitioners know what a defensible case requires. We’re inviting a small number of firms to help shape Ratioc ahead of a wider release, under NDA and on their terms.
We're working with a small number of firms ahead of a wider release.
If your compliance or surveillance team is evaluating AI tooling and data residency isn't negotiable, we'd like to talk. It's under NDA, on your terms, and there's no commercial obligation to start.
This page stays general on purpose. Once we're under NDA, we walk design partners through the full architecture, the deployment model, and how it would run inside your environment.
Prefer email? access@ratioc.ai, or find us on LinkedIn.
When you contact us through this site, we collect your name, email address, and any message you choose to send, solely to respond to your enquiry and discuss the design partner programme. We do not sell this information or use it for marketing unrelated to your enquiry. To run the site and handle enquiries we use a small number of service providers that process data on our behalf: Google (Apps Script and Sheets, which receive and store enquiries), Vercel (hosting and site analytics), Umami (site analytics) and, if you choose to book a briefing, Calendly (which handles the details you enter there under its own privacy policy). We may offer you the Arabic version of this site based on your browser language or approximate location; your approximate country is derived by our hosting provider from your IP address as the page loads, and we do not store it. You can ask us to delete your details at any time by emailing access@ratioc.ai.