We (a team of code-programming experts and many other teams) were tasked with training the AI model by correcting its responses, evaluating which response is better, and sending the results back to the AI model so it could learn from its mistakes. We used to create complex prompts, which were targeted to expand the model's perception limits to what the user might expect from it.

Generate a complex prompt

Evaluate model's the response

Critique model's the response

Rewrite the response to go above and beyond the user's prompt

Submit the results back to the model

Mostly yes, but there were moments related to management and pay that were kind of frustrating and led some people to leave. I liked how I could expand my proficiency in different fields while correcting the model's response and researching more about the problem.