Posts

Hello world!

ml@home ML tooling is in a really weird place right now. At $DAYJOB we use Google Cloud’s Vertex AI tools which works just okay. It does the job but it doesn’t really feel cohesive? So after setting up a homelab kubernetes cluster, I explored the open-source ML tooling landscape. Here are some initial notes. Seldon Core For ML serving. Deployed but unused right now. Feast A feature store. Could be useful for low latency inference with tabular models like recommender systems. Haven’t used it yet. The helm chart didn’t work for me, I had to point kustomize at an install.yaml to deploy the operator. ...