# Opportunity for AI people

> How to become rich and famous (not clickbait)

Published: 2020/12/21 • Author: Surya Dantuluri • Canonical: https://sdan.net/blog/opportunity-for-ai-people

I’ve been thinking of this strategy for a while after I used it once and got pretty good success out of it:

There’s this twitter account that posts pretty interesting papers; most of them have code open sourced on GH: [https://twitter.com/ak92501](https://twitter.com/ak92501)

But most of the time they’re hard to reproduce and use via Google Colab (colab imo is a mess, too much material ui which feels clunky)

A few months ago I built a system called Superscale, which makes it dead simple to: **create, deploy, and scale these AI models onto multiple GPUs/multiple nodes. It has processed hundreds of millions of jobs to date**. It does it with a pretty complex mashup of: VPNs, load balancers, container orchastraters, job queues, etc. However, all the python code is total trash; and ive gone in a bad habit of not updating to git local changes.... (and have no ci/cd); but it does the job: I was able to create an unrelased project called Cloakview (makes it impossible to detect faces by running an adversarial model against the image to change a few pixels here and there) within a day and scale it in real time (run on multiple gpu workers). And add a basic frontend that talks to very unsecure APIs in the backend.

Anyways, that’s not quite the point: if you can create a fast moving system that allows you to **clone and deploy** these models made by researchers within hours like I did with Superscale, in such a way that end users can easily interact with it (like i did with [https://twitter.com/sdan_io/status/1285687026386444288?s=20](https://twitter.com/sdan_io/status/1285687026386444288?s=20)) you can easily start generating money by adding a freemium model ontop of these models to go brrr to your audience and stripe account.

I’ve always wanted to do this, and might; but realistically can’t. Given that I just need a little motivation to clean up code and add some simple ci/cd, lmk if you’d want to possibly help; we can figure out some revenue share if it comes to that.