Overview
Sureva is a Swiss venture in the Surental that helps
Gemeinden, Vereine, and KMU adopt AI without surrendering data sovereignty.
The public face is IT and KI consulting (DSG-konform, on-premise, local presence
in Büron). Behind that sits a shared technical capability we are building:
self-hosted inference (Qwen, vLLM multi-LoRA, LiteLLM gateway) so advice turns
into systems that run inside the client's infrastructure.
I joined as Geschäftspartner and IT specialist: architecture, platform
roadmap, and the implementation layer that complements Sven's consulting and
local market anchoring.
The gap we are targeting
Swiss organizations want AI, but US hyperscaler cloud often conflicts with DSG /
nDSG expectations and internal governance. Typical integrators either resell
foreign cloud or stop at strategy decks. Sureva occupies the middle: **local
consulting plus hands that deploy and operate sovereign KI on-premise**.
What exists today
- Public marketing live at sureva.ch (DE): Private
KI & Chatbot, IT Beratung, workshops, Souveräne Cloud positioning.
- GitHub org Sureva-ch with venture
documentation repos and Ayentic consolidated under the umbrella.
- Platform roadmap documented (GX10 local sandbox, vLLM + LiteLLM production
path, RunPod burst capacity). Hardware procurement and first production
deployments are the next milestones, not retroactive claims.
My direct involvement
- Defined the umbrella architecture: Sureva (client-facing consulting) /
sovereign hosting layer / internal AI/backend stack, without exposing backstage
complexity to customers.
- Set up repo-per-venture documentation under Sureva-ch so each service has
a canonical GitHub home; vault notes mirror strategy for agents and portfolio
distillation.
- Own the self-hosted platform design (modular roadmap: base models, LoRA
adapters, gateway auth/rate limits, scaling policy) aligned with how I run my
own secondbrain and cost-aware LLM routing elsewhere on this site.
- Partner with Sven on narrative and delivery: he anchors B2G/local relationships;
I turn strategy into runnable infra and engineering decisions.
Technical direction (high level)
No private hostnames or client names here; those stay in internal runbooks until
explicitly cleared for publication.
Impact (early stage)
This is an active build, not a retrospective case study. Early outcomes are
structural: a credible public brand, documented platform path, and an org ready
for first on-prem KI pilots. Measurable client impact will land in this document
as milestones close (first Gemeinde/KMU deployment, production vLLM stack, GX10
online).
Learn more
- Venture site: sureva.ch
- Related writing on this portfolio: Sureva: building AI that survives contact with reality, Building the PublicBrain pipeline, Cutting LLM costs without losing capability