Keys and Caches: Vibe profile your ML models to get max performance.
Hey, Product Hunt! đ
Weâre Steven and Emilio, and we built Keys & Caches for anyone who has ever gotten lost in a sea of profiler GUIs and terminal windows just to figure out why their AI model is slow.
We love powerful tools like NVIDIA’s Nsight, but the steep learning curve, setup overhead, and the mental gymnastics needed to map low-level CUDA kernels back to our Python code was draining our productivity.
Our goal was to create something with the simplicity of torch.profiler but the depth of Nsight.
Keys & Caches is our open-source library that takes 2 lines of code to set up and gives you a single, unified view of your entire ML pipeline – from high-level Python code right down to the metal. It automatically connects the dots so you can spend less time debugging and more time building.
It’s entirely open-source and self-hosted, so your code and data always stay with you.
We’d be thrilled for you to check it out on GitHub (give us a star â if you like it!) and tell us what you think. We’ll be here all day answering every question.
Happy hunting!
Steven & Emilio
