# grait — Mariusz Kreft

> grait is the AI consulting practice of Mariusz Kreft (Berlin & Poland): AI strategy, self-improving agent workflows, Odoo-based solutions and private AI infrastructure — advised, built and run in production. Every engagement starts with an AI diagnosis of one process. Motto: Advise. Build. Run. Improve.

Key facts:
- Mariusz Kreft: computer scientist, product leader and entrepreneur with more than 15 years in software, AI and digital transformation. German, English, Polish (fluent). Based in Berlin and Poland.
- Current roles: Founder & AI Systems Architect, Machine.Machine (machinemachine.ai), since 2023; AI Consultant & Tech Partner, grait.io (agentic AI systems and digitization for a PE/VC firm and a Swiss manufacturing company); AI Engineering Trainer, Markant (Switzerland), current.
- Operated in-house: an 8-node NVIDIA DGX Spark cluster and 4× NVIDIA RTX 3090 GPU servers, all self-hosted.
- Agent fleet (Machine.Machine): 22 AI agents running; each has its own desktop, browser, tools and memory.
- Open source: GitHub org machine-machine, 134 repos.
- Workflow automation: previously n8n; now custom workflows run by herds of AI agents, orchestrated with m2herd.
- Contact: hi@grait.io · LinkedIn: https://www.linkedin.com/in/makecoolthingsreal
- Languages of the site: English (https://grait.io/) and German (https://grait.io/de/ — Deutsch: „Beraten. Bauen. Betreiben. Verbessern.“).

## Who it's for
- PE & VC operating partners: find where AI pays off in a portfolio company, then get it built and running (current work for a PE/VC firm).
- Manufacturers & Mittelstand: Odoo as the business backbone, AI agents for the repeat work, private AI on hardware you control (current work for a Swiss manufacturing company).
- Engineering teams: hands-on AI engineering training for senior developers (current recurring training at Markant, Switzerland).

## Services
- [AI strategy & architecture](https://grait.io/#service-strategy): use-case map ranked by value and feasibility, target architecture (cloud, private or hybrid), build / buy / run decision and roadmap, C-level presentation.
- [Self-improving agent workflows](https://grait.io/#service-workflows): custom workflows run by herds of AI agents that capture lessons and improve their own workflows, with human approval where it matters and an audit-proof (signed, hash-chained) decision trail. Can work on top of Odoo.
- [Odoo-based solutions](https://grait.io/#service-odoo): Odoo ERP as the business backbone, extended with custom addons and wired to AI agents that do the repeat work — e.g. a recruiting stack, dealflow, ERP for SMEs.
- [Private AI infrastructure](https://grait.io/#service-infrastructure): self-hosted LLM inference (vLLM / SGLang) behind an OpenAI-compatible API on hardware you control, from a single GPU server to a multi-node cluster; monitoring, runbooks, ongoing operation if wanted.
- [AI engineering training](https://grait.io/#service-training): hands-on training on agentic coding, orchestration, memory and evaluation for senior developers and engineering leads; workshops and C-level presentations for leadership.

## How an engagement starts
- [AI diagnosis](https://grait.io/#diagnosis): 1) Diagnose — you write a few lines about one process, we talk it through on a call, you get a short written recommendation: what to automate, where it should run, what to do first. 2) Build — the workflow, Odoo extension or private AI setup from the recommendation, built with your team. 3) Run & improve — run in production, and keep improving how it runs.

## Proof
- Current engagements: a PE/VC firm and a Swiss manufacturing company (agentic AI systems and digitization); Markant, Switzerland (recurring AI engineering training for senior developers).
- Earlier clients (appBauer): Procter & Gamble, Breitling, ADAC, Telekom. Transformation programs for corporates in retail, automotive, banking and pharma (freelance).
- [herdr-factory-loop-skill (m2herd)](https://github.com/machine-machine/herdr-factory-loop-skill): herds of AI coding agents through herdr in a spec-driven factory loop; m2herd evolve turns failed runs into reviewed lessons.
- [fleet-playbook](https://github.com/machine-machine/fleet-playbook): operating model for autonomous AI organisations.
- [openclaw-m2-memory-skill](https://github.com/machine-machine/openclaw-m2-memory-skill): agent long-term memory (Qdrant + BGE-M3).
- [ask-fable-skill](https://github.com/machine-machine/ask-fable-skill): delegates hyper-complex tasks to a Claude Code worker inside herdr.
- [All repos](https://github.com/machine-machine)

## Point of view
- [Recursive self-improvement for operating businesses](https://grait.io/rsi/): agent fleets that capture their own lessons, then review and amend their own workflows while humans keep the governance. Mariusz Kreft experiments with this in the Machine.Machine fleet.

## Request an AI diagnosis
- Email: hi@grait.io — subject "AI diagnosis request" (or "AI diagnosis — <topic>").
- Useful fields: your role; company type (PE/VC portfolio / manufacturer / retail / other); the process that should run itself; systems today (ERP/Odoo, tools); team size; where the data must live (cloud / EU / CH / own servers / don't know); timeline; preferred language (DE/EN/PL).
- Alternative: message Mariusz Kreft on LinkedIn (https://www.linkedin.com/in/makecoolthingsreal).
- Page: https://grait.io/#contact

## Links
- [Website](https://grait.io/)
- [CV](https://grait.io/cv/)
- [Deutsche Version / German version](https://grait.io/de/): KI-Beratung, KI-Diagnose, Leistungen und FAQ auf Deutsch
- [Lebenslauf (German CV)](https://grait.io/de/cv/)
- [Rekursive Selbstverbesserung (German RSI long read)](https://grait.io/de/rsi/)
- [LinkedIn](https://www.linkedin.com/in/makecoolthingsreal)
- [GitHub: machine-machine](https://github.com/machine-machine)
- [Machine.Machine](https://machinemachine.ai)

## Optional
- [Full text incl. FAQ, RSI long read and CV](https://grait.io/llms-full.txt)
- [Sitemap](https://grait.io/sitemap.xml)