Robert Mwangi

Key Skills

Software

Data Annotation Tech

Top Subject Matter

No subject matter listed

Top Data Types

Text

Top Label Types

Segmentation

Freelancer Overview

I am a dedicated AI enthusiast with hands-on experience in data labeling, annotation, and model fine-tuning across NLP and computer vision projects. I have worked extensively with tools like Labelbox and CVAT to annotate text, image, and video data, ensuring high accuracy and consistency—demonstrated by achieving a 98% inter-annotator agreement in sentiment analysis tasks. My technical toolkit includes Python, SQL, Jupyter Notebook, and Google Colab, which I use for data processing, cleaning, and validation. I have contributed to fine-tuning LLMs and image recognition models, improving model performance through refined annotation guidelines and thorough dataset preparation. I am passionate about delivering high-quality training data, following detailed guidelines, and supporting bias-free AI development.

Labeling Experience

data annotation

Data Annotation TechTextSegmentation

Project Title: Instructional Dialogue Annotation for Assistant-Style LLM Objective: To create a high-quality, diverse dataset of instruction-response pairs to fine-tune a large language model (LLM) for helpful, harmless, and accurate conversational AI. Scope of Work: Data Categorization: Annotate raw text prompts into categories (e.g., Creative Writing, Technical Q&A, Reasoning, Customer Service, Ethical Dilemma). Rewriting & Enhancement: Transform simple queries into well-structured, clear instructions. Example: Input: "tell me about python." Annotated Instruction: "Explain the Python programming language to a beginner, covering its main uses and key features in 3-4 concise paragraphs." Safety & Quality Labeling: Flag prompts containing: Harmful, unethical, or biased content. Requests for illegal activities. Factually incorrect premises (to be noted for model training). Style Specification: Tag the desired tone for the response (e.g., Formal, Friendly, Concise, Persuasive,

Project Title: Instructional Dialogue Annotation for Assistant-Style LLM Objective: To create a high-quality, diverse dataset of instruction-response pairs to fine-tune a large language model (LLM) for helpful, harmless, and accurate conversational AI. Scope of Work: Data Categorization: Annotate raw text prompts into categories (e.g., Creative Writing, Technical Q&A, Reasoning, Customer Service, Ethical Dilemma). Rewriting & Enhancement: Transform simple queries into well-structured, clear instructions. Example: Input: "tell me about python." Annotated Instruction: "Explain the Python programming language to a beginner, covering its main uses and key features in 3-4 concise paragraphs." Safety & Quality Labeling: Flag prompts containing: Harmful, unethical, or biased content. Requests for illegal activities. Factually incorrect premises (to be noted for model training). Style Specification: Tag the desired tone for the response (e.g., Formal, Friendly, Concise, Persuasive,

2023 - 2023

Education

Work History