Don't Kill a Fly With a Bazooka. Building AI That Actually Ships.
27 November 2026 · 09:15
Building has never been cheaper. You can ship five apps in a weekend. The catch is that four of them shouldn’t exist, and building the wrong thing at speed is still building the wrong thing.
The bottleneck has moved. It used to be “can I build this?” Now it’s “should I?” Eva-Maria Hempe, who leads NVIDIA’s Public Sector business across EMEA, unpacks the choices that decide whether anything you build gets used, not the ones that make a good demo.
You’ll leave with a sharper way to spot the right problem before you’ve burned your cycles on it, a test for when your problem is parallel but you’re still solving it one step at a time, and a practical view on why modularity, open models, and portability now beat raw capability. Plus the uncomfortable part: design partners, not speed, are the best early signal you’re onto something real.
And because a great build nobody can reach still fails, she makes the case for treating go-to-market as part of the build, not an afterthought. The fastest path to your users usually runs through the ecosystem already serving them: the partners, platforms, and communities you plug into rather than the sales motion you try to invent from scratch. Ecosystem isn’t a channel you bolt on at the end. It’s a strategy you design in from day one.
Building fast is a solved problem. This talk is about building right, and getting it to the people who need it.