You bring the ideas. I enhance and build them.
Open source, published · Anthropic courses · Production, not prototypes · Based in Stockholm
What I do for you
I build with the same toolkit I use on my own systems: agentic orchestration, RAG — search across your own data — and schema-validated responses.
Where the line goes
I map how the work actually gets done, exceptions included: what the human does, what the model does, and what no one should let it near.
What it gets told
An agent is only as good as its instructions. Prompts, context and tools get built against your data and your systems — not a demo that breaks on the first real question.
Catching what looks right
The dangerous answer isn't the one that crashes. It's the one that looks right and is wrong. You get criteria for an acceptable answer and a way to measure it.
Keeping it reliable
Guardrails, failure handling and an audit trail at every step: who approved what, on what basis. If you need it, I build the rest — frontend, backend, database, security.
What I'm building right now
All of it is built and running: my own agentic systems, plus two open-source developer tools.
Agentic AI Systems — Own tools · agentic AI
The RAG system behind my own work: a memory that ties everything I build together, multi-agent orchestration with schema-validated responses, and an autonomous work loop with review gates.
Claude Code, MCP, Multi-agent, RAG
Brink — Open source · developer tool
A Claude Code plugin that pauses an AI coding session just before it hits a usage limit, then resumes where it left off.
Node, Claude Code, npm, Hooks
MRC / Mobile Remote Control — Open source · developer tool
Start and steer local AI coding sessions from your phone — an agent in any codebase, from anywhere.
PowerShell, Claude Code, Windows, Linux
A little about me
I'm Linus Pisano, a self-taught FDE (Forward Deployed Engineer) based in Stockholm. I focus on integrating AI into existing systems and making it useful in people's day-to-day work.
My professional background is in CAD and fire safety. I currently work as a CAD drafter at seQRus Brand och Utrymningsteknik AB. Alongside that work, I have developed my skills in software development and AI through my own projects, hands-on experience and courses from Anthropic. I build AI agents, automations and systems that make information easier to find and use.
What drives me is the chance to free up time and make complex tasks more manageable. I want to help people understand where AI can contribute and give them solutions they feel confident using.
That is why my work starts with understanding the business: how people work, where their time goes and which problems are worth solving. From there, I build on the systems already in place, with clear boundaries for what AI should do and where human judgement is needed. The goal is to create solutions that can be understood, evaluated and developed over time.
Location: Stockholm, Sweden · Focus: Full-stack + agentic AI · Engagement: Freelance · Languages: Swedish, English · Status: Open to conversations
CAD & fire safety: AutoCAD, ProgeCAD, 3D modelling, Evacuation plans, SBA control drawings, SBF 1021:3, ISO 7010 · Full-stack: React, TypeScript, Vite, Tailwind, Flask / Python, Supabase, PostgreSQL, Row-Level Security, Playwright · Agentic AI: Multi-agent orchestration, Model Context Protocol, RAG, Schema-validated agent output, Guardrails & audit trail, Prompt & context engineering, Claude Code, Codex, AI workflow automation · Tools & production: Obsidian, Notion, Slack, Blender, CapCut, Higgsfield, fal.ai, Figma, Git & GitHub, Vercel, ngrok, FFmpeg
Self-taught — and still studying.
Everything I know is self-taught, apart from Anthropic's official courses.
- Claude Code 101
- Claude 101
- Claude Platform 101
- Introduction to Claude Cowork
- Claude Code in Action
- AI Fluency: Framework & Foundations
Questions I get asked
If yours is not here, ask it on the call.
Do I need to know what I want automated before we talk?
No. Most people know something takes too long, not which part a machine should handle. Working that out is the first thing we do.
Where does my data go?
Nowhere we have not agreed on first. Which model provider is used, what is sent to it, and what stays inside your systems is written down before anything is built. If data would leave the EU, you hear it first.
Do I have to replace our current systems?
No — close to the opposite. Everything is built on top of what you already run, connected to your existing data and tools. Migrations are expensive, risky and usually unnecessary.
What if AI is the wrong answer for us?
Then I say so, and you have lost a conversation rather than a budget. Plenty of problems are better solved with a spreadsheet, a fixed process or one small script.
How long does something like this take?
A first working version usually takes days, not months, because it sits on what you already have. The time goes into agreeing what a good answer looks like and making it reliable enough to trust.
Can we do this in English?
Yes. I work in English and Swedish. Documentation comes in whichever your team actually reads.
Let's build something.
An idea, a problem, or a question about what AI could do for you? Send it over.
Emails answered within 24 hours.