Runs on your infrastructure
On-premises or private-cloud deployment only. No inference call, log line, or fine-tuning pass ever leaves the perimeter you define.
A holding company for applied AI
We build artificial intelligence with one condition:
it stays where you put it.
inlokus designs and deploys AI systems for institutions that operate under real constraints: regulatory, contractual, reputational. Every product we build runs on infrastructure you control, answers only from data you provide, and leaves nothing behind.
- 100%
- On-premises or private-cloud deployment
- Zero
- Client data retained or trained on after delivery
- Res Ipsa
- Our first product, built for litigation teams
Approach
I.
II.
III.
Three commitments that apply to everything we ship.
Nothing is retained
We do not train on client data, store it beyond the session, or reuse it across engagements. What comes in is not kept.
Built for one domain at a time
We do not ship a general assistant with a vertical skin. Each product is built around the workflows, terminology, and failure modes of a single regulated field.
Products
A small group of focused products, not one platform stretched thin.
Legal · e-Discovery
Res Ipsa
An air-gapped case wiki. Res Ipsa ingests raw discovery files and builds an interlinked, fully cited knowledge base for litigation teams; every claim traceable to its source page, nothing exposed to the open internet.
We're scoping additional verticals with the same discipline: likely finance, healthcare, and public-sector records. Each one ships as its own product, not a feature.
Compliance
We design around the standards you already answer to.
-
Self-hosted or private-cloud deployment, chosen by the client; never a shared multi-tenant service.
-
No training on client data, under any circumstance, at any stage of an engagement.
-
Every generated output carries a traceable reference back to its source.
-
Every deployment is reviewed against the client's existing compliance framework before go-live.
Contact
Tell us what you can't risk getting wrong.
One conversation is usually enough to know whether we're a fit.