Junior AI / software engineer · relocating to Germany
I scope the problem, build it, and talk to the people who'll use it while I do. Right now I'm building and leading a GenAI document pipeline at a German fintech.
currently: shipping BriefPilot sprint 2 · learning German
field: shipped
problem:Asset onboarding at vountain (a fintech in Hanover) ran on manual data entry.
decision:Proposed and now lead a GenAI document pipeline: automated extraction with validation and a human check before anything is trusted.
outcome:Automated extraction across 0+ documents a week.
adoption:Running in production at vountain, replacing manual entry.
problem:Untracked async standups, and client Slack requests turned into tickets by hand.
decision:n8n + LLM automations: a standup summariser reporting to leadership, and a Slack-to-ClickUp intake with deadlines.
outcome:Daily progress visibility without meetings, and client messages become tracked tasks automatically.
adoption:Company-wide at two teams.
problem:A Dubai school needed a learning platform built.
decision:Delivered it over a six-month freelance engagement.
outcome:Promoted mid-project from building it to reviewing the junior developers' pull requests.
field: work
Photograph a German official letter and it reads the text back to you. That is the first step toward telling you, in plain language, what it means and what to do.
The OCR pipeline and a photo quality-gate (it asks you to retake an unreadable scan instead of showing garbled text) are shipped. Classification and the plain-English explanation come next. Next.js · FastAPI · Postgres · Docker · CI.
Quantum states are hard to reason about as vectors of complex numbers. This is a visual, interactive representation that makes gate effects and entanglement observable: build qubits, apply gates, and watch a Bell pair and a teleportation protocol play out in real time.
The state engine is fully typed and separated from the rendering layer. Every push builds and deploys automatically to the live site.
field: how_i_work_with_ai
It's my job title, so I'll be specific about it. Generating code is the easy part now; knowing what not to hand over is the job.
Boilerplate, first-draft tests, and mechanical refactors: work where I already know the answer and just need it typed out.
Architecture, and the boundaries between parts. In BriefPilot the AI provider sits behind one interface I designed, so a model can be swapped without touching anything else.
I read every diff, and mypy --strict plus CI catch what I miss. When the model wasn't sure what a document was, I made it return nothing rather than a confident guess.
field: decisions
Provider-Agnostic LLM Architecture Under a Zero-Budget Constraint
field: path
Freelance developer
Shipping real things for real clients
Business analyst
Learning to scope a system before building it
Project coordinator
Shipping with a team, on deadlines
PM & automation
Building the tools I used to spec
Data & GenAI engineer, vountain
Where I am now
Every step added the next skill. I went from writing the specifications to building the systems those specifications described.
field: contact
I'm looking for a Werkstudent or junior AI / software engineering role in Germany, remote now and on-site once I relocate.