Aapo Tanskanen

I actually do hands-on work too, not just pontificate and banter around. That means figuring out why the pipeline fell over on a Friday afternoon, training models, and architecting real AI-powered software products.

It’s easy to get pulled into every news flash about a new AI model and its benchmark scores. Thinking beyond that means asking a different question: what can we actually build with this today? A service that didn’t exist last year. A team process that used to take a week and now takes an afternoon. The model is the least interesting part.

Taking a strategy that’s been sitting in a slide deck and actually building the thing. The gap between "we should use AI for this" and a working system in production is where most projects die. That’s the gap I close.

I trained dozens of Finnish LLMs from scratch years before ChatGPT was released, so I have deep hands-on experience in building and using these kinds of models. I tend to push back on overly complex, trendy AI solutions that try to solve a problem we’ve already solved with something simpler.

Having advised global clients like Mozilla, BMW, and Rightmove on AI, I stopped assuming the serious expertise lived abroad. Our work in Finland held up next to all of them. That reset how I think about where good AI actually gets built.