Hacker News Reader

Related: Solid Project

- HN discussion: https://news.ycombinator.com/item?id=25989698

- A great article about the project: https://ruben.verborgh.org/blog/2020/12/07/a-data-ecosystem-...

Slightly related, we're working with a similar philosophy. As a machine learning consultancy that has done many learning projects for enterprise, we're building our machine learning operations, "MLOps", platform (https://iko.ai) to simplify our work. However, what we're doing is working from the architecture level to have as little and preferrably no sensitive information on our service. We're architecting it so that you give us specific access to deploy on your cluster, and everything happens there: the notebook servers are there, your data is where you choose to put it, your training jobs are there, your experiments are tracked there. Your models are deployed there.

I have a saying that the platform should be able to run on a Raspberry PI.

One of my personal pet peeves working with the team is to be able to disappear without impacting them, and it has become the same with our platform: it must be able to disappear users having to scramble to exfiltrate or export their work or data from our infrastructure, because it simply is not there.

Reply on news.ycombinator.com