OVERVIEW

Knowing what’s on the screen is just not enough! Our clients wanted their model to not just “read” images, but they wanted their model to work. like a rational thinking human. They wanted their model to analyze, reason and interpret the data that was given on the screen. Understanding their requirements, Soul AI got to work

Approach

We deployed 100+ experts with very strong profiles (e.g., top universities, flawless English), plus a background in fields related to data interpretation, such as economics, finance, mathematics. With this cherry-picked team, the model is being trained right now as you read this.

The project is underway with 3 types of tasks:

1. Review image > Write a detailed prompt
‍
2.Review image > Writing a detailed prompt and a detailed response
‍
3.Review image + review prompt > Reviewing/modifying the model response

Reach out to us at hey@soulhq.ai for more information, work samples, etc.

Multi-Modal SFT

OVERVIEW

Approach

other work