A new age for South Africa

Private AI

The enterprises running AI on hardware they own are pulling ahead. The ones still renting are one outage, one export rule, one price hike behind. Built in Cape Town for South African teams shipping AI under POPIA, data-residency, and cost constraints. Pick the right side.

Talk to us

Infrastructure

Private AI compute you own. NVIDIA DGX Spark systems and workstations, on-prem or hybrid. Specified, installed, and maintained for you.

Private Lab

A private ML lab on hardware you own: a workbench for experimentation, fine-tuning, embeddings, and evaluation, with MLOps on your local stack.

AI Engineering

AI in production: internal platforms, agents, RAG and knowledge systems, observability, workflow automation, and system integration.

AI Guardrails

Policy enforcement for developer AI tools and agents, across IDEs, repositories, CI/CD, MCP servers, and model gateways.

Flagship products

Built and run on the same stack.

RapportIQ

AI for relationships

Tessa Van Wyk

Three follow-ups, no reply

At risk

James Coetzee

Asking about renewal terms

Warm

Aisha Khan

Two paid invoices, fast

Engaged

Tessa Van Wyk

Sentiment trending down · 9d silent

AI recommendation

Pattern matches 73% of churn risks last quarter. Call her before Friday.

RetainUpsellRe-engage

Relationship intelligence

RapportIQ

Explore relationship intelligence

Tells you which customers need attention, why, and what to do next. Sentiment, risk, lifecycle phase, and next-best-action across WhatsApp and email.

Naledi Sithole

Contact-center platform

NPU Agent

Explore private call centers

A private call center rebuilt as an AI platform. WhatsApp, voice, web chat, email, identity verification, and a single admin shell to run the floor.

.wow · session

$/feature add user-auth

✓feature.md created

$/plan

✓7-step plan written

$/build

✓claude takes the keyboard

$/memory save

✓context preserved

Way of working

Agentic WoW

Explore the right way to work

A feature-driven way of working with a compounding memory loop: Feature, Plan, Build, Memory.

A private AI computer and a laptop showing training metrics on a warm wooden desk in a Cape Town home, Table Mountain visible through the window.

Control and convenience

A private AI lab that fits on a desk.

A quiet box on your desk that answers the moment you ask. Nothing leaves the room. And because we build the harness around it, the evals and the guardrails, you know it learned from your data and behaves the way you decided. That is the quiet confidence of an AI that is yours.

See the hardware we build on

Coming soon

Meet

Neuro

A private memory and context platform for AI agents.

Source-aware memory, governed retrieval, and durable handovers for Codex, Claude Code, Copilot, and your own agents. A single context gateway across repositories, services, sessions, and machines.

5y+Long-horizon retention. Five-year-plus archives without a single re-prompt.

Get on the early-access listSelf-host or run on your NPU Labs node.

In active developmentNPU Labs Research

Built on

PostgreSQLQdrantNeo4jRedisForgejoVault

Capabilities

Context Gateway
Librarian Agent
Semantic + graph retrieval
Provenance + lineage
Layered retention
MCP across agents

Live agent queryGateway online

agent.query("how does our auth flow work?")

Context Gateway routing

Memory3 records · 5y, 2y, 7d

Decision from 2021: passwordless rollout reasoning still load-bearing.

Sourcesrc/auth.ts · main@a4f9c1

Live code, freshness 12 min.

LineageADR-014 · confirmed by reviewer

Provenance traced. No conflicts.

MCP across all agents

How we work

We build it, then we hand it over.

We spec the hardware for your workload, build the lab, fine-tune the models, and write the guardrails. Then we hand you the box and the runbook. From that point it runs on your terms, in your building, on your schedule.

An NPU Labs engineer handing a private AI computer and a deployment runbook to a client with a handshake, Cape Town visible through the window.

Build the AI stack you actually own.

Take back control. The data and the engineers were always yours, and so was the right to decide how AI runs inside the business, but somewhere along the way that became someone else's call. We design the infrastructure, build the lab, ship the agents, and write the guardrails so the stack ends up exactly where it should have been all along. Yours.

Build with us