> visitor: "Can I talk to Einstein?"
agent.voice.connect(persona)
portrait.style("anime") → QR ✓ Generative AI Experience Hub for Flagship Stores
From Electron prototype to multi-store web app with voice agents & generative portraits.
002_AGENTS.MCP
Chatbots that know your docs. Agents that do the boring work. AI features your users actually use. Prototyped in days, production-ready in weeks.
Building an impressive AI demo takes an afternoon.
Getting it to work on Monday morning, with real users, messy data and a CFO asking about token costs? That’s the actual job.
That’s the part I love. ❤️
✅ AI features inside your product: smart search, assistants, summaries, generation.
✅ AI agents that read emails, tickets or documents and put the results where they belong.
✅ MCP servers that give AI assistants safe, typed access to your systems.
✅ Automations in n8n your team can see and change themselves.
Day 1: We pick one use case. Not ten.
Week 1: You click through a first working version on your real data.
Weeks 2–6: Weekly iterations. Then honest numbers, and a clear decision: scale it, hand it over, or kill it cheaply.
Voice agents and AI portraits for a telco’s flagship stores. An MCP-powered concierge. A live AI art installation.
Different problems. Same approach: ship early, learn fast.
> visitor: "Can I talk to Einstein?"
agent.voice.connect(persona)
portrait.style("anime") → QR ✓ From Electron prototype to multi-store web app with voice agents & generative portraits.
tool: compare_routes
in: { start, destination }
out: 🚗 18 min · 🚲 14 min · 🚶 41 min A typed MCP server + n8n agent on Gemini that compares car, bike & walking routes from plain language.
A physical prompting station feeding a live Stable Diffusion pipeline built in ComfyUI.
A first working AI prototype usually exists within days. A full sprint with real users and honest numbers takes 4–6 weeks.
An open standard that lets AI assistants use your tools and data through one interface. Build one MCP server for your system, and Claude, ChatGPT or n8n can all use it.
Whatever fits: Anthropic Claude, OpenAI, Google Gemini or open-weight models hosted in the EU. Model-agnostic, so switching stays cheap.
It's part of the architecture from day one: EU hosting or self-hosted models where needed, data minimisation, no training on your data. I work alongside your DPO.
Also on the menu
042_NEXT_STEP.SH
30 minutes. No sales script. You tell me what you want to build. I tell you honestly how fast we can get there, and what it takes.
Prefer to chat first? Send me a DM on LinkedIn or email me.