Moonglow (YC S24) connects local Jupyter notebooks to remote compute providers.
Most modern ML research starts in Jupyter notebooks, which are great for exploration. But when it’s time to scale up, they hit a wall. The infrastructure just isn’t there to easily connect notebooks to serious GPU power.
This means that ML researchers spend valuable time converting their notebooks into scripts, a time-consuming and slow process. Then they SSH to a remote machine and copy their environment and all that data over. Or, they use hacky workarounds like remote Jupyter servers, which are error-prone, or cloud compute notebook providers, like Google Colab, which only provide a limited GPU selection.
Founded by Leila C. and Trevor Chow, Moonglow lets researchers use the cloud GPUs of their choice, right from their notebook editor. With just a click of a button, you can move a Jupyter notebook from a CPU to an H100 in less than a minute. By automating the DevOps, Moonglow lets researchers focus on what matters: the ML problems themselves.
Try it out at https://moonglow.ai.
Congrats on the launch, Leila and Trevor!
🚀 https://lnkd.in/gx26WWbM