|b| Expertise areas

Three practices.
One shared rigour.

Design, industrialise and operate the technology platforms that carry your strategic priorities — under control, at the expected level of performance.

AI · LLM · ML · Vision

Artificial intelligence

Language models, machine learning and vision systems — designed, industrialised and deployed on sovereign, cloud or hybrid infrastructure.

Stack: PyTorch · vLLM · Transformers
Deployment: On-prem · private GPU · Cloud
Governance: OPA · PCL · audit trail
Metric: p95 latency ≤ 800 ms

Web · Critical applications

Web development

Critical applications and high-traffic interfaces built on React, Next.js, Astro and TypeScript — audited for performance and accessibility.

Stack: React · Next.js · Astro · TS
Performance: Lighthouse 95+ / WCAG 2.1 AA
CI/CD: E2E tests · SAST · atomic rollback
SLA: 99.95% per contract

Cloud · Operation under control

Cloud & infrastructure

Controlled migration to AWS, Azure or Google Cloud, Kubernetes orchestration and measured reduction of total cost of ownership.

Cloud: AWS · Azure · GCP · OVHcloud
Orchestration: Kubernetes · Terraform · ArgoCD
Observability: OpenTelemetry · Grafana · Loki
TCO: -30% on average over 12 months

|b| Intervention principles

Four rules, always held.

P-01

Engineering first

Code reviewed, tested, documented. Architectures justified, decisions traced — nothing that won't hold under a peer's scrutiny.

P-02

Verifiable sovereignty

Your infrastructure, your keys, your data. We never take ownership of what must remain under your control.

P-03

Numbers over promises

Contracted targets (latency, availability, TCO), published measurements, argued adjustments.

P-04

Continuity assured

Beyond delivery: operation, maintenance, evolution. Service contract calibrated, no break on the horizon.

|b| Next step

A project to frame?

Share your context. Our engineers get back to you within 24 working hours with a first argued reading — no commitment.

Discuss your project