What I build,
ship and train.

What we do at Kamoo: building agents that stay up in production, rolling out on-prem AI on a DGX Spark for data that can't touch the cloud, and training teams to do this themselves.

A

AI that keeps your data in the building

Local models for customer data that can't go to the cloud, and stays out of US jurisdiction. Hardware advice, vLLM tuning, quantization, eval. Including handover to your ops team.

Per project · on request
B

An agent that keeps running in production

MCP servers, eval suites, retrieval, tool-calling. Cloud (Anthropic, OpenAI) or local on your own hardware. From prototype to production deploy with monitoring and eval gates. No demos that only work on a Friday.

Per project · on request
C

Your team builds this themselves

One-day via Kamoo, or in-company tailored to engineering teams. Hands-on with your own codebase: agents, MCP, eval, on-prem choices. No slides about what AI is, but something concrete to take back with you.

In-company · on request

This track applies to custom projects (A and B). For training you enrol via the training page.

01

First talk, free

Week 0

One no-strings talk where we break down what you want to automate and what role agents or a local chatbot could play. Sometimes AI turns out not to be the answer, sometimes it's simpler than you thought. No quote afterwards, just an honest read.

02

Eval-first

Week 1 – 2

Before a single line of code: an eval suite that measures whether it works. Sounds boring, prevents 80% of the pain I see crop up elsewhere.

03

Building, short loops

Week 2 – 8

Weekly demo, eval per release. What passes the eval goes live. What doesn't, doesn't. Not even if it looks good.

04

Handover

Week 8 – 10

Your team takes over. Documentation, runbooks, on-call training. I stay in the chat for questions for up to four weeks.

Esc