AI is easy in a demo. Harder at the border.
I work on AI in real industry — automotive, logistics, research — and on the messy layer between tools and outcomes.
I show where AI survives real industry work.
First stamp: machines. Next border: software that has to move them.
Engineer First
Mechanical engineer first. I learned what wrong systems do before I asked AI for answers.
M.Sc. at the University of Stuttgart: robotics, simulation, control. There, a bad assumption shows up as a machine doing the wrong thing. In AI, it hides longer.
Kinematics, trajectory planning, MATLAB/Simulink
Multibody dynamics, MSC Adams, FEM
SolidWorks, CATIA, AutoCAD
System dynamics, hardware interfacing
Theory can enter the country. Production decides whether it may stay.
After The Demo
Four industries. Same question each time: where does automation still create value after the demo ends?
Data must travel
I map where data gets stuck: documents, systems, owners.
Tools need tasks
I match AI tools to real workflows, not trends.
Demos must survive
I test whether use cases hold in daily work.
Each industry stamped the passport differently. The startup stamped it hardest.
My Failed Startup
ADHOK Intelligence: AI for company communication, founded with my best friend in London. Pitched on stage two days after the idea existed. One paying customer. Closed.
Two hundred cold calls later I knew the market better than our own roadmap. Wrong order. The stranger who joined us after that first pitch became my closest friend in Germany — which says more than the pitch deck ever did.
Team beats idea
The idea changed three times. The people decided everything.
Discovery before product
Two hundred calls taught more than six months of building.
One customer is data, not validation
A signed contract is one stamp, not permission to scale.
Failure stamped the passport harder than success would have. Next: build something boring enough to survive.
Built Real Systems
A registered student association in Stuttgart, built from zero, run as president.
Constitution, finances, events — then the part I enjoyed most: a self-built registration and payment system in Google Apps Script. Real money, zero budget. It runs because the boring path came first.
0
registered members
0 €
software budget
1
payment system in production
A system for people taught me the same lesson factories teach: if it depends on memory, it breaks.
Test Before Hardware
Digital twins: test the factory in software before hardware makes mistakes expensive.
Virtual commissioning — faults found in the model, not on the shop floor. Current work: parallel discrete-event simulation and real-time partitioning for hardware-in-the-loop. Master thesis from October 2026, same domain.
VIBN: validating plant behavior before the shop floor sees it
Parallel discrete-event simulation and synchronization protocols
Splitting models so hardware-in-the-loop can keep time
Research only earns its stamp when someone outside the lab can use it. So I publish what I learn.
What Works
AI across industries, explained for people whose time is expensive. Short, specific.
What I learn in real process work, published while it is fresh. One problem, one usable thing per piece.
AI in automotive — an insider view
What survives outside the demo
ChatGPT vs Claude vs Copilot
Pick tasks, not sides
I failed my AI startup
The scar tissue, told straight
Every stamp so far was mine. The next checkpoint is yours.
Your crossing
You know AI matters. The useful question is where it gives you leverage. That border is my work — and for me there are no borders, only possibilities.
One line is enough. I read this before anything else.
Find where AI pays rent in your company.
Three hours: data infrastructure audited, three automatable processes named, one-page roadmap. German or English.
Or send the border directly
avalad.bw@gmail.comThe passport does not move at 92 percent. The rest is a step.