Zakayo Dennis

Key Skills

Software

AWS SageMaker

Anno-Mage

Appen

Argilla

CloudFactory

Data Annotation Tech

Datatroniq

Deep Systems

Google Cloud Vertex AI

Humanatic

Labelbox

LabelImg

Mighty AI

Mindrift

OneForma

Redbrick AI

Remotasks

Scale AI

Toloka

Telus

V7 Labs

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Geospatial Tiled Imagery

Image

Top Task Types

Bounding Box

Computer Programming Coding

Mapping

Prompt Response Writing SFT

Translation Localization

Freelancer Overview

I have over three years of experience in AI training data annotation, specializing in computer vision data labeling for self-driving cars, object detection, and machine learning models. My expertise includes bounding box annotation, polygon segmentation, keypoint tracking, and video frame-by-frame labeling, ensuring high-quality datasets for AI development. I am highly skilled in using annotation tools like Labelbox, CVAT, Scale AI, and V7 Labs, and I understand the importance of precision, consistency, and quality control in AI training datasets. In addition to computer vision, I have LLM evaluation and text generation expertise in English, Swahili, German, and Spanish, helping refine AI-generated responses for accuracy and relevance. My ability to work with multimodal AI data—including text, images, and videos—sets me apart. I bring a strong analytical mindset, attention to detail, and a deep understanding of AI training workflows, making me a valuable asset for any AI-driven project.

Labeling Experience

OTS Dataset Program - Red Teaming Domain - en_US

OneformaTextText GenerationEvaluation Rating

The OTS Dataset Program - Red Teaming Domain focuses on evaluating and improving the safety, reliability, and ethical alignment of Large Language Models (LLMs). The project involves testing AI-generated responses by comparing and rating them based on criteria such as helpfulness, honesty, and harmlessness. The goal is to identify potential weaknesses, biases, and risks in AI-generated text to enhance model performance and ensure responsible AI deployment.

The OTS Dataset Program - Red Teaming Domain focuses on evaluating and improving the safety, reliability, and ethical alignment of Large Language Models (LLMs). The project involves testing AI-generated responses by comparing and rating them based on criteria such as helpfulness, honesty, and harmlessness. The goal is to identify potential weaknesses, biases, and risks in AI-generated text to enhance model performance and ensure responsible AI deployment.

2024 - 2024

Video Annotation

LabelboxVideoBounding BoxSegmentation

This project focused on image and video annotation to enhance the accuracy and efficiency of AI models, particularly in computer vision applications. As annotators we were responsible for labeling images and videos using techniques such as bounding boxes, segmentation, keypoint annotation, and object detection to train machine learning models effectively.

This project focused on image and video annotation to enhance the accuracy and efficiency of AI models, particularly in computer vision applications. As annotators we were responsible for labeling images and videos using techniques such as bounding boxes, segmentation, keypoint annotation, and object detection to train machine learning models effectively.

2024 - 2024

Fireweed - Falcon LLM Multi-Lingual

AppenTextTranslation LocalizationRLHF

Project Fireweed focuses on the development and fine-tuning of a Large Language Model (LLM) capable of operating across multiple language domains. The goal is to enhance the model's versatility, ensuring it can accurately understand and generate text specific to each domain, while maintaining contextual and linguistic integrity. The project will involve domain-specific data collection, model adaptation, and evaluation to create a robust, scalable solution for diverse language applications. Key tasks include data preprocessing, model training, performance evaluation, and deployment for real-world use cases.

Project Fireweed focuses on the development and fine-tuning of a Large Language Model (LLM) capable of operating across multiple language domains. The goal is to enhance the model's versatility, ensuring it can accurately understand and generate text specific to each domain, while maintaining contextual and linguistic integrity. The project will involve domain-specific data collection, model adaptation, and evaluation to create a robust, scalable solution for diverse language applications. Key tasks include data preprocessing, model training, performance evaluation, and deployment for real-world use cases.

2024 - 2024

Flamingo Code Sql

Scale AIComputer Code ProgrammingText GenerationText Summarization

This project focuses on SQL database management, query optimization, and data processing to enhance the efficiency and reliability of structured data systems. Participants will work with Flamingo Code SQL, leveraging SQL to handle database operations, optimize queries, and analyze data for various applications.

This project focuses on SQL database management, query optimization, and data processing to enhance the efficiency and reliability of structured data systems. Participants will work with Flamingo Code SQL, leveraging SQL to handle database operations, optimize queries, and analyze data for various applications.

2024 - 2024

Artemis Annotator - General Domain - English

OneformaTextText SummarizationRLHF

This project involves evaluating AI-generated responses by assessing their Harmlessness, Honesty, and Helpfulness to ensure high-quality AI training data. As an evaluator, I compared two responses to a given prompt, determine which is better, and assign ratings based on specific criteria. The goal is to refine AI models by improving their ability to generate accurate, ethical, and useful responses.

This project involves evaluating AI-generated responses by assessing their Harmlessness, Honesty, and Helpfulness to ensure high-quality AI training data. As an evaluator, I compared two responses to a given prompt, determine which is better, and assign ratings based on specific criteria. The goal is to refine AI models by improving their ability to generate accurate, ethical, and useful responses.

2024 - 2024

Education

Work History