AI Project Manager. AI/ML engineer by training.
I started in AI engineering, but over time I found myself spending more time on the parts around the model: clarifying the problem, keeping delivery moving, checking whether an output is actually reliable, and deciding when a human needs to step in.
That is still the kind of work I enjoy. AI can draft. Engineers can improve it. But someone still has to decide whether the result is good enough to use.
AI draft → engineering review → evaluation → human decision → ship
I do not treat a successful output as a correct output.
I make the decision owner visible, especially when the system is handling something people may rely on.
And I choose the architecture based on the risk. Sometimes the right answer is a carefully constrained workflow. Sometimes it means deliberately not using GenAI.

