Air-gapped sovereign LLM platform, MCP multi-agent automation and token FinOps
Production AI that the CISO approved and the CFO welcomed, with every byte of client data kept in its jurisdiction.
At a glance
Client
A regulated European and MENA financial services institution.
Team
1 principal AI and data architect, 2 sovereign ML and agent engineers and 1 cloud and PQC security engineer
Duration
10 weeks
The challenge
The client wanted to automate complex bilingual (English and Arabic) KYC document extraction, compliance auditing and internal knowledge retrieval. Strict data residency laws, the MENA personal data protection laws (PDPL), the EU AI Act and DORA, prohibited sending raw customer personal data to public US consumer LLM APIs.
Its first cloud AI prototype also suffered from runaway token inference costs and the risk of hallucination.
What we built and connected
Private inference with zero data egress
A private inference cluster for large and small language models inside a sovereign regional cloud VPC, with permission-aware GraphRAG that enforces role-based access to documents.
Supervised MCP agents
Autonomous agents on the Model Context Protocol (MCP), supervised by people on the loop, that extract, cross-check and post KYC, trade-finance and onboarding documents directly into core systems (intelligent document processing, IDP 2.0).
A token FinOps gateway
Semantic caching and dynamic model routing that send 75% of routine tasks to quantised small language models and keep frontier models for complex reasoning.
Systems and stack
- AI and machine learning
- Mistral
Measured results
- lower monthly AI token and GPU inference spend, through semantic caching and small-model routing
- 52%
- faster KYC and document onboarding, with deterministic citation links for human compliance officers
- 78%
- and zero external leakage of personal data, passing internal CISO and regulatory risk audits on the first review
- 100% data residency
XSIS gave us production-grade sovereign AI that our CISO approved and our CFO applauded — cutting our inference bill in half while keeping every byte of client data inside our jurisdiction.
Start with a conversation, not a contract.
Tell us what you need to build, connect or secure. An architect goes through it with you, then recommends where to start.
What happens next
Agreeing a time
Our team emails you to agree a time for the session.
Pre-consulting session
A free 30-minute online pre-consulting session: we go through your goals, systems and constraints.
Readiness review
In a free 48-hour readiness review, a Principal Architect checks your project against the 6-point Definition of Ready and recommends the track that fits.
Choosing the track
Scope-ready projects receive a binding proposal within 3–5 business days. Concept-stage projects start in the Blueprint Studio. A live system that is failing, undocumented or needs integration starts with a diagnostic of 1–2 weeks. The Blueprint Studio and the diagnostic each have a fixed fee, quoted in the proposal.