Subscribe
Sign in
Start new thread
p/mlblocks
A no-code Machine Learning solution. Made by teenagers.
Visit Product
Sarvasv Kulpati
mlblocks — A no-code Machine Learning solution. Made by teenagers.
Featured
3
•

As students in high school, we realised that even though it is amazing, machine learning is really difficult to implement into projects.

To remove the hassle we faced ourselves, we created MLBlocks.

-Train models with only 50 images per class

-Create them without expensive GPUs

-Deploy on any platform through our API

-NO CODE REQUIRED

Replies
Best
Sarvasv Kulpati
Maker
👋Hi PH!👋 Sarvasv from the mlblocks team here.👶 As two teenagers really interested in AI🤖, we see how amazing it would be if it was used in new and innovative projects. But because of the technical difficulty, powerful computers, and large amounts of data needed, it seems like many people avoid it. To solve this we made 🙌mlblocks🙌. Here’s a couple of things we help with: -Create ML models with only 50 images per class -Train models on the cloud without powerful computers -Access your ML models from any platform through our REST API -No code required This is our first version of the product (and the first real project we’ve ever launched), so we hope you can check it out and give us some feedback!🚀🚀
Alex Wawl
@sarvasvkulpati Hi guys, congrats with launch. Few questions: Are you training models on GPUs in the cloud? If your service is free what happens if I upload 10Gb or 100Gb images? How do you store uploaded data? Can u use it on your own cases? Don't you think that 50 images per class is nothing? I can host my models on your servers. Will it always be free or you will add payments later?
Sarvasv Kulpati
Maker
@alexwawl Hello Mr. Wawl, These are all great questions! 1. For now, we managed to get access to some GPUs at no cost, which allows us to train peoples models on them 2. We do have limits to the number of images and classes allowed, though we could think about scaling as soon as we nail the product 3. We store uploaded data on aws, though I'm not entirely sure what you mean by can you use it on your own cases 4.Yes! 50 images per class is really low, we created our backend so that you need the fewest training images possible, and 50 is the lowest we could get with reasonable accuracy 5. We're definitely looking into monetisation, though at the moment, it's completely free so that we get as much feedback as possible Thanks for asking!