AI and DATA at the service of your business.

JAIKIN is an AI, automation and custom-software consultancy for SMEs and mid-caps, with an industrial anchor in Alsace; we work across France. We work in three phases: on-site mapping of AI opportunities, rapid prototyping to validate value before investment, then leading the production deployment with our expert partners.

Our engagements are scoped project by project — the cost depends on the scope (targeted diagnosis, prototype, full industrialization) and the sensitivity of the data involved. We don't publish a catalog of offers: every engagement starts with a free 30-minute conversation to understand your context, followed by a tailored quote.

The use-case mapping runs over 2 to 4 weeks with a quantified deliverable. A working prototype takes 4 to 8 weeks depending on complexity. Production industrialization depends on your IT landscape and the criticality of the use case — we frame the timeline upfront based on a validated POC.

Yes — we work across France, remotely for ongoing follow-up and on-site for key phases. Our industrial anchor is in Alsace / Grand Est (head office in Fegersheim, office in Paris) and we work in French, German and English, but our reach is nationwide: we support SMEs and mid-caps wherever the use case matches our expertise (production, industrial IP, sensitive commercial data).

We design architectures compliant with the EU AI Act, favoring sovereign deployment infrastructures. Typical stack: context-adapted LLMs (Claude, GPT-4, fine-tuned open-source models), RAG, open-source orchestrators, IT system integration via APIs. Our in-house tools are accelerators at the service of consulting — not standardized products.

Three phases structure our intervention. One: understand — on-site diagnosis, quantified mapping of priority use cases. Two: validate — rapid prototyping to demonstrate value before industrialization. Three: deploy — production rollout with our expert partners and full integration into your IT system. You keep a dedicated point of contact from end to end.

Our Responsible AI Charter v1.0 (April 2026) rests on two structural pillars. One: data anonymization and minimization by default — systematic pseudonymization before any transmission to models (LLMs, embeddings, vision), hosting exclusively within the European Union (Scaleway, OVH, Heroku EU), retention strictly limited to project needs and a right-to-erasure mechanism embedded in every POC. Two: open source and sovereignty preference — Mistral, Llama and Qwen models run locally via Ollama or vLLM for low-risk use cases, European providers prioritized at functional equivalence, fully reversible architectures (no vendor lock-in, JSON/CSV/ONNX/Parquet exports). Six operational principles complete these pillars: human-in-the-loop on every high-impact decision, explainability (documented prompts, cited RAG sources, exposed confidence scores), active bias analysis, AI Act and GDPR compliance with a dedicated compliance lead per engagement, digital sobriety (SLMs prioritized, caching and batching, low-carbon hosting), full transparency on AI components, their limits and risks. Public charter, annual review, whistleblowing channel at victor@jaikin.eu with a 15-day response window.