Joseph Kama

Key Skills

Software

CVAT

Labelbox

Label Studio

OneForma

Internal/Proprietary Tooling

Appen

Top Subject Matter

LLM evaluation in French

e-commerce

map

Top Data Types

Audio

Document

Text

Top Label Types

Bounding Box

Emotion Recognition

Evaluation Rating

Prompt Response Writing SFT

Translation Localization

Freelancer Overview

I'm a dedicated Language Specialist and AI Trainer with over three years of experience. I specialize in translation, annotation, and data labeling for Artificial Intelligence and Machine Learning applications. As a freelancer for leading Language Service Providers, I contributed to the training of renowned models and agents that drive AI innovation. My expertise includes prompt and response annotation for Large Language Models (LLMs) and AI agents, leveraging Reinforcement Learning from Human Feedback (RLHF) and Supervised Fine-Tuning (SFT). Additionally, I perform comprehensive data annotation across text, image, and audio formats, contributing to robust AI training datasets. I leverage my strong time management, clear communication, and keen attention to detail to consistently deliver top-quality work, even under tight deadlines. Let’s connect to explore how I can contribute to your AI and language projects!

Labeling Experience

Pdf Annotation

Label StudioTextBounding Box

Contributed to a high-precision project to structure PDF documents through detailed annotation for enhanced machine readability. Created and labeled bounding boxes for text, images, and other elements (e.g., section headers, list items, captions) using tools like Label Studio, ensuring accurate categorization and transcription of content.

Contributed to a high-precision project to structure PDF documents through detailed annotation for enhanced machine readability. Created and labeled bounding boxes for text, images, and other elements (e.g., section headers, list items, captions) using tools like Label Studio, ensuring accurate categorization and transcription of content.

2024 - 2024

General Preference Collection for LLM Response Evaluation

Internal Proprietary ToolingTextFine TuningEvaluation Rating

Contributed to a large-scale project to improve large language model (LLM) performance by evaluating and ranking responses based on Harmless, Honest, and Helpful (HHH) dimensions. Analyzed prompts and paired responses, assessing their quality across HHH criteria and assigning Ranking Quality Scores (1–7) to determine preference. Ensured responses were safe, accurate, and aligned with user intent, adhering to strict guidelines on sensitive topics, factual accuracy, and tone. Conducted quick web searches to verify information when needed, maintaining high standards for objectivity and precision.

Contributed to a large-scale project to improve large language model (LLM) performance by evaluating and ranking responses based on Harmless, Honest, and Helpful (HHH) dimensions. Analyzed prompts and paired responses, assessing their quality across HHH criteria and assigning Ranking Quality Scores (1–7) to determine preference. Ensured responses were safe, accurate, and aligned with user intent, adhering to strict guidelines on sensitive topics, factual accuracy, and tone. Conducted quick web searches to verify information when needed, maintaining high standards for objectivity and precision.

2023 - 2024

Bryant French ASR Project

AppenAudioSegmentation

French transcription project of short audios. The audios are from famous social media, candidates were required to work as annotator and QA checkers to do transcription and segmentation.

French transcription project of short audios. The audios are from famous social media, candidates were required to work as annotator and QA checkers to do transcription and segmentation.

2021 - 2022

Education

Work History