Hacker News new | comments | ask | show | jobs | submit login
Ask HN: Are any startups working on Distributed ML?
6 points by hsikka 4 months ago | hide | past | web | favorite | 7 comments
I've been thinking of writing an SDK for distributed ML processes like federated learning, would it make sense as an enterprise case?

Amazon SageMaker attempts to address this problem. I have also read some google papers about using tens or hundreds of GPUs to solve problems like with alphago. The issue with most enterprises is that they do not know how to get value out of ML. And they won’t need distributed ML until they can make money with ML.

My experience is with pitching cable companies on ML projects.

Tensorflow has a multi-gpu sdk that I used with sagemaker.

Oasis labs is.

White paper: https://arxiv.org/pdf/1804.05141.pdf Their site: https://www.oasislabs.com/

Note that it's built around blockchain.

Problem looking for solution

I think you meant "solution looking for a problem..."

Can you elaborate on this?

Engineers like to work on cool technology and then try to find a problem that it solves which can be different from solving real problems for users.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact