Mercy Kamau

Key Skills

Software

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Text

Top Label Types

Text Generation

Freelancer Overview

I have extensive experience in high-quality data labelling and RLHF (Reinforcement Learning from Human Feedback), specifically focusing on Natural Language Processing and sentiment analysis. My work involves transforming raw, unstructured data into high-signal training sets by applying rigorous annotation guidelines and maintaining a consistently high inter-annotator agreement (IAA) score. I have successfully contributed to projects involving named entity recognition (NER) and intent classification, ensuring that the ground truth data is free from bias and edge-case errors that could compromise model performance. What sets me apart is my dual focus on speed and semantic accuracy. Beyond just labelling, I have experience in auditing datasets for quality assurance and providing feedback to refine labelling taxonomies. I am proficient with tools like Labelbox and CVAT, and I possess a deep understanding of the iterative nature of model training, which allows me to anticipate the needs of data scientists and deliver datasets that require minimal cleaning.

Labeling Experience

RLHF & Model Safety (Best for Generalist/Writing Roles)

Scale AITextText Generation

Performed Reinforcement Learning from Human Feedback (RLHF) to align LLMs with human intent. Tasks included ranking model responses based on truthfulness, helpfulness, and safety. I identified subtle hallucinations and ensured compliance with strict ethical guidelines to prevent biased outputs.

Performed Reinforcement Learning from Human Feedback (RLHF) to align LLMs with human intent. Tasks included ranking model responses based on truthfulness, helpfulness, and safety. I identified subtle hallucinations and ensured compliance with strict ethical guidelines to prevent biased outputs.

2023 - 2024

Education

Work History