Kokkola, Finland — 63.8°N
Clear software,
that makes work calmer.
I'm Jari Urpilainen. I build products that use AI, coordinate technology projects, and teach people to use intelligent tools — without the hype.
Three kinds of work, one belief: technology should ask less of people, not more.
AI-assisted products
Software where artificial intelligence removes work instead of adding features — drafting, summarising, and quietly disappearing into the workflow.
Technology projects
Coordination from funding application to final report. EU frameworks, stakeholders, budgets — the unglamorous machinery that makes good ideas actually land.
People who can use both
Training that turns AI skeptics into capable, critical users. Taught in the learner's own professional language, never in tech abstractions.
Projects are stories: a problem, a solution, and what it changed.
An unusual route to AI — and why every step still matters.
Electronics taught systems. Logistics taught flow. Teaching taught clarity. Nothing was wasted.
- 01 · The beginning
Electronics
Learning how things actually work.
Circuits don't accept hand-waving. A cold solder joint fails whether or not your theory was elegant. Electronics taught me to think in systems, to debug methodically, and to respect the difference between understanding something and almost understanding it.
Carried forwardSystematic debugging. Respect for the physical layer.
- 02 · Flow
Logistics
Systems are only as good as their weakest handoff.
Logistics is invisible when it works and catastrophic when it doesn't. Moving real things through real constraints taught me that optimisation is about bottlenecks, that information quality beats information quantity, and that most process problems are communication problems wearing a disguise.
Carried forwardBottleneck thinking. Processes as products.
- 03 · The toolset
ICT
Software is leverage over everything else.
Moving into ICT connected the earlier worlds: the systems thinking of electronics, the flow thinking of logistics, now expressed in code and infrastructure. I learned that the hard part of technology is rarely the technology — it's fitting it to humans.
Carried forwardFull-stack perspective. Technology in service of workflow.
- 04 · Transmission
Teaching
You understand something when you can teach it to a skeptic.
Teaching vocational students is a masterclass in honesty. Abstractions bounce off; relevance lands. Explaining technology to people who need it for work — not for exams — permanently changed how I design, write, and build. Clarity became a professional standard, not a nicety.
Carried forwardRadical clarity. Patience with real learning curves.
- 05 · Orchestration
Project Management
Someone has to make the whole thing land.
Coordinating projects — budgets, stakeholders, EU frameworks, deadlines — taught me the unglamorous machinery behind every good idea. Vision without administration dies in month three. I learned to love documentation, decision logs, and the quiet craft of keeping many people moving in one direction.
Carried forwardDelivery discipline. Stakeholder empathy.
- 06 · The multiplier
Artificial Intelligence
Everything before this was preparation.
AI arrived as the technology that touches every previous chapter: it debugs like electronics, flows like logistics, builds like ICT, explains like teaching, and coordinates like project management. I work with AI daily — building products with it, teaching it, and studying where it genuinely helps versus where it merely impresses.
Carried forwardPractical AI judgment. Building with, not just about.
- 07 · Now
ICTU
Bringing it all into one room.
ICTU is where the whole path converges: an education-sector AI project that needs teaching skill, project discipline, software craft, and technology judgment simultaneously. It's also proof of a belief — that the most valuable AI work right now is helping real institutions and real people put it to use.
Carried forwardThe synthesis. And the next chapter being written.
Four working principles, tested in classrooms and codebases.
AI is a multiplier, not a mind.
It amplifies the judgment you bring to it. Expertise in, leverage out. Vagueness in, confident nonsense out.
Trust comes from traceability.
An AI feature earns adoption when its output can be checked — citations, sources, visible reasoning. Never ask users to believe.
The best AI is quiet.
Intelligence should reduce the moments technology demands from a human. If a feature adds notifications, it isn't finished.
Adoption is a teaching problem.
The gap between what models can do and what organisations use is not technical. It closes one relevant, honest demonstration at a time.
Writing that treats the reader as smart and busy.
AI won't replace you. Refusing to learn it might.
The honest version of the AI-and-jobs conversation, from someone who teaches it to people whose jobs are supposedly at risk.
7 minApproachableTeaching AI to skeptics
What vocational classrooms taught me about AI adoption that no product launch ever did.
9 minApproachableA small team's AI stack, honestly reviewed
What I actually use to build software as a near-solo developer in 2026 — including what I stopped using and why.
11 minIntermediate
Got a problem worth solving?
No forms, no funnels. Write me an email like you'd write to a colleague — I read every one, and I answer in person.
Usually replies within a day. Finnish or English — molemmat käyvät.