Analytical databases
Self-hosted ClickHouse and Postgres tuned for analytics — sub-second queries on your own metal, at a fraction of warehouse-as-a-service pricing.
What we build
The cloud, minus the part where they keep your data.
Private AI inference
Open-weight LLMs served from your own hardware — chat, embeddings, and document intelligence with zero data leaving the building.
Workflow & app platforms
Self-hosted orchestration and backends — n8n, Supabase, and the plumbing between them — deployed as one coherent, maintainable stack.
Secure exposure
Zero-trust access to internal tools — tunneled, authenticated, loopback-bound by default. Public only what must be public.
0 %
data residency — every byte on hardware you own, in a room you can point to.
0
per-seat SaaS licenses — open-source stacks with no meters running while you sleep.
0
documents in one private semantic index we run in production — search and RAG without a cloud API in sight.
How an engagement runs
Designed like infrastructure, documented like a product.
Scope
What must stay private, what it costs you today, and what hardware it actually needs — sized honestly, not aspirationally.
Architect
The full diagram before anything is installed: services, network boundaries, backup and recovery, upgrade paths.
Deploy
Containerized, reproducible installs with secrets managed properly and every port justified.
Hand over
Runbooks, monitoring, and a rebuild-from-zero recipe — your team operates it; we stay on call if you want us.
You leave with
- A production stack you own outright — no vendor can price-hike or sunset it.
- Runbooks and a disaster-recovery recipe tested by actually running it.
- A security posture worth auditing — zero-trust access, loopback-bound services, least privilege.
Start
Tell us what can't leave the building.
One short brief. A reply within two working days. NDAs welcome — expected, actually.