AboutAbout me
I’m Gabriele, a senior PM at Dynatrace building agentic AI for observability.
I care more about whether things actually work than whether they sound impressive, which I’ve found to be a useful disposition in an industry where the noise is deafening, and almost nobody stops to ask if the thing holds up.
The interesting work is in the specifics: what the models can actually be trusted with, where they’re confidently wrong, what users will tolerate, what they won’t notice, and what they’ll quietly work around when no one’s looking. That last one tells you the most: people route around what they don’t trust, and they almost never say so.
I came to product through industrial design, then taught myself the tech. No CS degree, no coding background, just a lot of stubbornness, long days and late nights, and a habit of pestering the experts around me until it clicks.
Building AI products has made me a far more critical user of them. Once you’ve worked on the context, the tools, the orchestration, the evals, the guardrails – you stop being impressed by the surface and start digging into what’s underneath. The magic gets a lot less magical and a lot more interesting.
Design school taught me to sit with a problem before reaching for a solution, and to treat simplicity as something you earn, not something you start with. Product rewards exactly that. The rest, no one taught me: showing up curious, keeping things candid, and the discipline to follow through once the hype wears off.
I’m Gabi. I build agentic AI at Dynatrace, which mostly means figuring out what LLMs can actually be trusted to do, and what just looks great in a demo but falls apart in production.
I have very little patience for a lack of integrity or authenticity, and AI is running low on both right now. Barely anyone stops to ask if any of it actually holds up. I’d rather talk about what works – or, more honestly, what doesn’t.
The good stuff is always in the specifics: where the models are confidently, gloriously wrong, what users will learn to put up with, what they won’t even notice, and what they’ll quietly work around the second you stop looking. That last one tells you the most – people route around what they don’t trust, and they’ll almost never admit it.
I didn’t get here the conventional way. Industrial design in South Africa 🇿🇦, then a software job in a country I’d just moved to 🇦🇹 (wtf, right), which meant teaching myself the tech (and the German) to keep up – no CS degree, no coding background, just a lot of stubbornness, a lot of coffee, and a habit of pestering the smartest people around me until it clicks.
Building this stuff has made me thoroughly sceptical of it all, in the best way. Once you’ve been elbow-deep in the context, the system prompts, the tools, the evals, the licensing (yes, really), and the guardrails (omg, don’t get me started on guardrails), the magic wears off fast – and what’s underneath turns out to be far more interesting anyway.
Design school taught me to sit with a problem until I actually get it, instead of rushing to a solution – and that simplicity is earned and taste is learned. Turns out product management runs on exactly the same principles. The rest, no one taught me: showing up curious, calling bullshit when I see it, and actually following through once the hype’s worn off and everyone’s wandered off to the next shiny thing.
Get in touch
Connect on LinkedIn or send me an email.