Hello, world

I've spent fifteen years designing silicon — GPU schedulers, interconnects, memory subsystems, verification methodology — and the most interesting problems were never purely hardware or purely software. They lived in the space between the two: the decisions where an architectural choice in RTL changes what a compiler can do, and vice versa.

That's the space I want to write about here: AI hardware–software co-design. What it actually means in practice, what I've learned building chips at NVIDIA and Intel and now leading engineering at Modular, and where I think the industry is headed as AI workloads keep reshaping the hardware underneath them.

I won't promise a schedule — but I will promise substance. If any of that sounds interesting, stick around.

— Jenny