CyberIA Labs designs, secures and ships systems where data must be protected, augmented and proven. Consulting, architecture and development — from scoping to production.
CybersecurityBlockchainArtificial IntelligenceRarely isolated. Most often combined within the same project — and that is where the value sits.
Measure the real risk, harden the infrastructure and the code, assess compliance. An engineer's approach: documented and verifiable.
LLMs that make it past the demo: integrated into the business, kept under control, and never leaking your data.
Proof as a deliverable: integrity, traceability and legally enforceable value.
Using generative AI should never mean handing your documents to a third party. Depending on data sensitivity, three regimes — which often coexist within a single project.
The strongest models on the market, where the nature of the data allows it: verified usage terms, no reuse for training, upstream anonymisation and traffic segregation.
Models hosted under European jurisdiction with contractual guarantees, when data is sensitive but internal infrastructure cannot host the service.
Llama, Mistral and open models running on your own infrastructure or in a private environment. No data leaves your perimeter — the default for confidential documents.
Two decades at the core of telecom digital security: the understanding of hardware, networks and standards that makes the difference on industrial and connected projects.
Standardisation work with the GSMA: SM-SR / SM-DP definition and profile management.
GSM 2G / 3G / 4G / 5G architectures, SMSC, V2X, embedded connectivity and operator VPN.
eSE, SIM cards, Javacard OS, OTA administration and cryptographic library integration.
Security, contactless payment and embedded connectivity for consumer and industrial devices.
Three areas of expertise delivered by the same team: no hand-off to manage between security, data and proof.
Running generative models on confidential documents: local or private execution, segregation, access control and outbound traffic control.
Anchor input data, model versions and generated deliverables: who produced what, from which sources, and when.
Identity, signature, qualified timestamping and integrity proof aligned with ETSI / ANSSI frameworks.
An integrator's approach: start from real usage, validate what the technology genuinely brings, then deliver.
Understand the business before the technology. Qualify the expected gain, rule out what adds nothing, set measurable success criteria.
Technical design, security model, component selection and integration into the existing landscape. Nothing disposable.
Development, testing, hardening, containerised deployment and CI/CD pipelines.
Documentation, team training and full autonomy. You remain the owner of your solution.
Representative assignments, anonymised.
A travel agency held thousands of past quotes containing the essence of its know-how: itineraries, hotels, stages, services, day-by-day planning. A RAG architecture indexed that archive, then generated a complete proposal — dates, stages, hotels, flights, car rental, activities, day-by-day description — faithful to the agency's own practice. A significant part of the assignment addressed model limitations: never add days, never drop a flight or a rental, never alter reference services, never invent information.
An agent that leads the exchange rather than passively answering it: it follows the conversation, identifies what has already been gathered, asks for what is missing and maintains usable context. Collected data is then structured and turned into a finalised document, in a predefined format, ready for review and delivery to the client.
Technical leadership of a solution for enforceable emails and electronic registered mail, backed by blockchain and a qualified timestamping service. Full development of the timestamping component and of the ETSI / ANSSI documentation framework, through to PSHE certification.
A commercially deployed bottle traceability solution: NFC chip, mobile application, backend and NFT drop marketplace on Polygon. Solution architecture and full token lifecycle design, from bottling to resale.
Security audit assignments on networks and infrastructures, evaluation and program management of secure component certification — EMVCo, GSMA eSA, GlobalPlatform, Common Criteria — and infrastructure assessment for post-quantum readiness. Work carried out alongside product teams, R&D and evaluation laboratories.
EBIOS · hardening · infrastructure audit · cloud & IoT pentest · HSM/KMS · PKI · post-quantum · EMVCo, GSMA eSA, GlobalPlatform, Common Criteria
International, sovereign and self-hosted LLMs · RAG · autonomous agents · embeddings · Ollama, Llama, Mistral · LangChain · TensorFlow · vectorised DMS
Ethereum · Polygon · Tezos · Solana · Stellar · Bitcoin · smart contracts · notarisation · eIDAS timestamping · NFT
Docker · Kubernetes (AWS) · GitLab CI/CD · Vault · Java / Spring · Python · Node · eUICC / eSE · Javacard · OTA · 2G-5G
Our reading of the technical news.
Picking a “sovereign” model does not make a use case compliant. What can actually be controlled is which data leaves, what the provider does with it, and what trace you keep.
The standard has been published since 2024. What stalls enterprises is not picking an algorithm — it is knowing where cryptography actually sits, and how long already-deployed equipment will live.
Describe the context in a few lines. Reply within 2 business days, with an initial technical opinion — not a sales brochure.
