I am Nawinn, born 1984, based near Frankfurt am Main. I took my first PC apart as a child and started programming at sixteen. Professionally: 8 years as an application developer and server administrator across four IT companies — ICOM Computer (where I learned the fundamentals), CPC-IT, SDS Informationstechnik and a continuing-education center.
Stack back then: C#, C++, Java, HTML/CSS, JavaScript, interface programming (SOAP, XML, legacy protocols), Windows servers, Linux backend configuration, network setup. Not just code — also the infrastructure the code runs on.
Between 2011 and 2024 I worked outside of IT — but programming never went away. A steady stream of tools for personal use: IP scanners for the home network, RAM-dump utilities, WSL cloning scripts, backup automation, a home-lab and CPU monitor, a Wake-on-LAN utility and a server-monitoring bot. In parallel I followed the technical shifts closely: own VPS in Germany as a serious AWS alternative, local LLMs like Mistral Small 3.1, Astro, Flutter. Decisions that are my core stack today.
Since 2025 I work full-time as a developer again — with modern AI coding assistants as a constant co-pilot in the development workflow. Not a gimmick: a fundamentally different working mode. The AI takes boilerplate and typing overhead off my plate; architecture and judgement stay with me. Concretely: this website, a full Django backend with auth and multi-tenant logic, a Flutter app with iOS/Android builds — each in weeks instead of months.
Why "workshop"? Because software is a craft. Every project has its particularities. No industrial "sprint cycles" for one-person engagements — instead considered work steps, clean documentation, traceable decisions.
Why "stackschmiede"? Because I cover the full stack — from server setup to UI. Backend, frontend, infrastructure, increasingly AI integration with local models. A small developer network stands behind me for holiday cover and emergency backup.
I find it absurd how routinely personal data gets handed over today — geo-locations, contact graphs, photo content, payments, background profiling. This data feeds AI models that are increasingly deciding things about us: insurance rates, credit scores, maybe tomorrow's job ratings.
My answer: privately I have moved to my own solutions wherever possible. Own photo server, own drive, own messenger, local AI, own banking aggregation. Exactly those solutions I now offer to others. Cloud-exit is not abstinence — it is a different architecture.
For the first 3–5 reference projects I deliberately offer pilot terms: reduced fixed price in exchange for public reference mention after project end. Client internals stay confidential (NDA standard); only what you release gets referenced. After that, regular rates. Win-win: experienced full-stack work below market price, visible portfolio for me.
Before a fixed price is on the table, I sit down with you: goals, domain language, risks, existing systems. The outcome is a scope memo with time and cost framework — readable and re-negotiable. No blind-flight quotes.
Every week a preview build runs on a test domain — no PDF status reports. You see progress, not my interpretation of it. Feedback flows directly into the next iteration.
Source code sits in your GitHub/GitLab organisation from day 1. No black-box development, no vendor dependency. At project end you take over — or we keep working. Your choice, not mine.
Handover including admin docs, deployment runbook, backup concept. 30 days warranty for production bugs included. Optional: managed hosting or retainer. What if I drop out? The code lives with you, the docs live with you — any experienced Django developer picks up.
I write code the next person — or me in six months — can read without puzzling. Named variables, documented invariants, no clever tricks. Exactly what AI assistants were not built to produce on their own: clean, readable context.
AI coding assistants are my co-pilot in the development workflow, not my autopilot. Eight years as a C#/C++ application developer, interface programmer and server admin sit behind every architecture decision. The AI writes faster. The responsibility stays mine.
For your shipped product I deliberately choose on-prem LLMs (Mistral Small 3.1, Codestral) on your server — either on one of my workshop servers in Germany or on your own infrastructure. That is your business-continuity insurance: pricing, APIs and regions stay under your control. If you deliberately want a large cloud model (Claude, GPT) I integrate that via its API just the same — a conscious choice, not the default.
I host in Germany/EU, use open formats, build on widely-adopted standards. No Vercel dependency, no Firebase lock-in. Harder upfront sometimes — pays off the first time a vendor conflict hits.
Home base near Frankfurt am Main. On-site meetings in FFM, Darmstadt, Mainz, Wiesbaden, Cologne — I like taking the train.
Active today: Dart/Flutter (iOS+Android), Python/Django, Go, TypeScript, Astro, HTML/CSS.
Legacy competence: C# (.NET Framework 2.0–4.x, WinForms, ADO.NET), C++, Java, interface programming.
Infra: Linux server admin, Caddy/Nginx, Podman/Docker, Postgres.
One core developer, complemented by a network for holiday cover, firefighting and specialist tasks. Transparent communication about who is currently in the seat.
8 years as an application developer and server admin at IT companies. Then a sector pause with continuous tech work on the side. Full-time return in 2025 as an AI-augmented developer.
Client communication and documentation in German. English for technical docs and international repos.
Also no projects with obvious ad-tracking architecture, dark patterns or tracking-pixel farms.
Three channels, one contact. Reply within 24 hours on business days.