AI Training Contributor

In one sentence

A remote contractor role contributing to AI model training and evaluation for leading AI labs, via Mercor’s expert annotation platform.

What I did

Worked on AI training projects focused on improving how language models reason through complex problems. The work involved three main tasks: writing structured reasoning trajectories — step-by-step reasoning chains that models learn from — evaluating model outputs against defined quality criteria, and identifying where reasoning breaks down in code-based tasks.

The projects contributed training data to language model fine-tuning pipelines, specifically in areas of code reasoning and logical problem solving. This type of work sits within the post-training phase of model development, using supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF) techniques.

Outcome

Ongoing. Selected by Mercor as a domain expert contractor, contributing to active AI training projects for leading AI labs.