Building a platform where next-gen AI models get trained by humans!
Building a platform to train AI models using RLHF and SFT annotations!

When a AI model is build it build in stages, it starts from a general SFT means feeding data into it and then there comes RLHF where humans comes to tell AI if the result is right or wrong and by how much.
CONTEXT
Humans make AI smarter!
In training AI, humans play a big role by giving feedback that helps it learn, improve, and make smarter choices.

Onjective
Building a platform where AI get’s trained by humans!
For building a successful AI training platform we need to build 2 things...

Studio / Dashboard
Where user can see their current projects, tasks and task history along with training.

Annotation Platform
Where the annotators will be performing the tasks and actually training the AI.
Insights
We only need a Platform consist of these 4 things!
There was only few competitors in the market and most of them have locked design, and they don’t wanted to show their UI design due to confidential reasons.

Our product have 2 types of users!
First ones are annotators, who will be working on our project & Second one will be QCs, who will be checking the quality of the tasks done by the annotators and this can grow in future.



Annotator's View
MVP: Consisting of many fields and more than 3 pages, asking a lot of details that was not even required for the internal team.
1) Takes ~3 minutes to complete
2) Lack of hierarchy in the content
3) Capturing useless details

QC View for Tasks
MVP: Consisting of many fields and more than 3 pages, asking a lot of details that was not even required for the internal team.
1) Takes ~3 minutes to complete
2) Lack of hierarchy in the content
3) Capturing useless details
Annotation Tool
Platform where AI get's trained!!
There was only few competitors in the market and most of them have locked design, and they don’t wanted to show their UI design due to confidential reasons.
There are various use cases for annotations!
Use cases like Video annotation, Image annotation and Image annotation are some common type of annotations that we have catered to!

We tried to cover the most complex ones first!!
Although the tasks that we have from client was very simple but we actually have tried to build the complex ones in the designing so that we can build a much scalable platform!






Studio Design
Training users for tasking
Annotators need training pre-tasking or before starting to work on the project as training a model is something very personalised and custom therefore a Training page with courses & workshops !

Studio Design
Payments details for the tasks users do!
After every successful task, the task goes to the QC person and if they approved it.. user receive the payment. It can either be in Dollars or Rupees

Studio Design
Building a Robust Design System!!
Starting with Colors
Keeping things basic.. I created shades of the colors and keep it simple.

The most important part of design, Typography!!
Have sat with developers to define the breakpoints for typography and use variables to define it in Figma!

Sitting with Developers to figure out the Breakpoints for typography!!

Defining Breakpoints in Figma Variables & Modes for making design responsive easily!
Building & Defining Components
It wasn't perfect but as a solo designer I tried to make the best out of my time!

