Pattin Mugambi

Key Skills

Software

Mindrift

Lionbridge

Other

Top Subject Matter

No subject matter listed

Top Data Types

Text

Audio

Top Label Types

RLHF

Red Teaming

Prompt Response Writing SFT

Classification

Evaluation Rating

Data Collection

Audio Recording

Freelancer Overview

I am a detail-oriented data specialist with hands-on experience in data annotation, image labeling, text categorization, and semantic tagging for AI training projects. My background in linguistic analysis and telephone interpretation—especially in legal and medical domains—has given me a strong foundation in understanding semantic intent, cultural nuance, and tone. I excel at providing high-quality human-in-the-loop feedback, conducting adversarial testing to identify edge cases, and ensuring data integrity with over 98% accuracy. Proficient in Ubuntu Linux, Docker, and advanced web research tools, I am adept at handling complex guidelines and delivering precise results in fast-paced, project-based environments. My skills in quality assurance, search relevance, and technical literacy make me a valuable contributor to AI and NLP initiatives.

Labeling Experience

Search Query Intent & Content Relevance Evaluation

LionbridgeTextClassificationEvaluation Rating

Analyzed user search queries to evaluate the relevance of e-commerce and web results. Applied linguistic expertise to determine "User Intent" and graded the accuracy of search engine outputs. Ensured that results were culturally and linguistically appropriate for the Kenyan market.

Analyzed user search queries to evaluate the relevance of e-commerce and web results. Applied linguistic expertise to determine "User Intent" and graded the accuracy of search engine outputs. Ensured that results were culturally and linguistically appropriate for the Kenyan market.

2025

Adversarial Linguistic Evaluation & LLM Testing

MindriftTextRLHFRed Teaming

Acted as a controlled adversary to stress-test Large Language Models (LLMs). Engaged in complex, multi-turn dialogues to evaluate the model's emotional intelligence, instruction-following, and consistency. Developed unique user personas to identify logical inconsistencies and subtle failure modes in AI reasoning.

Acted as a controlled adversary to stress-test Large Language Models (LLMs). Engaged in complex, multi-turn dialogues to evaluate the model's emotional intelligence, instruction-following, and consistency. Developed unique user personas to identify logical inconsistencies and subtle failure modes in AI reasoning.

2025

Audio recording

OtherAudioData CollectionAudio Recording

Collected and annotated hyper-local environmental data, including acoustic noise mapping and geospatial point-of-interest (POI) imagery. Verified the accuracy of automated transcriptions and metadata for regional dialects and urban landmarks.

Collected and annotated hyper-local environmental data, including acoustic noise mapping and geospatial point-of-interest (POI) imagery. Verified the accuracy of automated transcriptions and metadata for regional dialects and urban landmarks.

2025 - 2025

Education

Work History