About feder8d
feder8d started from a simple observation: every organisation we talked to had data they wanted to use with AI — customer records, source code, clinical notes, legal work, internal IP — and zero appetite for sending it to a public cloud AI provider. Their compliance team, their legal team, or their customers had said no. So they were stuck: either ship a worse product, or take on regulatory risk they couldn’t justify.
feder8d is the third option. Open-weight models, run inside a private boundary in your jurisdiction, with an OpenAI-compatible API on top so your existing code keeps working. The source is published so your security team can read exactly how we handle your data — every line, every commit, every commit message. That transparency is a diligence credential, not a free tier: running it in production needs a commercial agreement with us.
Mission
Make private AI infrastructure boring — so the teams running regulated, confidential, or sovereign data can spend their attention on the experience instead of the threat model.
Principles
- No third-party AI in the prompt path. Open-weight models only, unless you explicitly bring your own provider key per collection.
- Self-host first. If we wouldn’t run it on our own infra, we don’t ship it.
- Region sovereignty matters. Your data resides where it belongs — we won’t move it without telling you.
- Third-party SaaS minimisation. Every external dependency is a sub-processor. We keep the list short and public.
- Inference is stateless. Prompts and completions are never persisted on the model side. Audit logs store hashes only.
- OpenAPI is the contract. One spec, generated SDKs, no surprises.
Where we're headquartered
feder8d is built in Cape Town, South Africa. Your data resides in af-south-1; inference runs in eu-central-1 for GPU availability — only tokens cross that boundary, never raw documents. The disclosure is in every DPA. See /legal/privacy.
Talk to us
General: hello@feder8d.com
Sales: sales@feder8d.com
Security: security@feder8d.com