Christabel Okoro

Key Skills

Software

SuperAnnotate

Other

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

Document

Image

Video

Top Task Types

Evaluation Rating

Object Detection

Prompt Response Writing SFT

RLHF

Text Generation

Freelancer Overview

I have hands-on experience working with AI training data, especially in large language model (LLM) projects. At Micro1, I annotate video content with a strong focus on accuracy, context, and alignment with project-specific training goals. I consistently manage high-volume, high-detail tasks, reviewing, validating, and maintaining consistency in complex annotation guidelines. My work directly supports machine learning development, and I actively contribute to improving workflows through quality assurance feedback and collaboration with team members. Previously at Turing, I helped train LLMs by crafting prompts, annotating datasets, and evaluating model outputs across RLHF, RAG, and SFT projects. I was recognized as a top performer for consistently exceeding quality targets. I also produced step-by-step trajectory recordings and provided QA feedback for fellow annotators.

Labeling Experience

CUA project

Internal Proprietary ToolingVideoPoint Key PointTracking

In this project, I worked on training data for AI agents designed to understand and perform tasks within software applications, similar to how a human would interact with a screen. Using a tool, my screen activity was recorded as I completed spreadsheet-based tasks like data entry, formatting, and navigation, as well as other daily applications, like browsers,mail,music,office and presentation software, etc. These recordings were then converted into screenshots, with metadata such as mouse clicks, drag paths, keyboard inputs, and cursor coordinates.

In this project, I worked on training data for AI agents designed to understand and perform tasks within software applications, similar to how a human would interact with a screen. Using a tool, my screen activity was recorded as I completed spreadsheet-based tasks like data entry, formatting, and navigation, as well as other daily applications, like browsers,mail,music,office and presentation software, etc. These recordings were then converted into screenshots, with metadata such as mouse clicks, drag paths, keyboard inputs, and cursor coordinates.

2025 - 2025

Video Annotation project

SuperannotateVideoObject DetectionAction Recognition

The project involved detailed frame-by-frame annotation of videos, with timestamps. Annotations were given for general descriptions, type of video, lighting, scene cuts, and camera descriptions. The quality adhered to involves the word count, as well as accuracy of the annotation.

The project involved detailed frame-by-frame annotation of videos, with timestamps. Annotations were given for general descriptions, type of video, lighting, scene cuts, and camera descriptions. The quality adhered to involves the word count, as well as accuracy of the annotation.

2025 - 2025

RLHF

Internal Proprietary ToolingImageRLHFEvaluation Rating

This project involved image tagging (text free/light and text rich), as well as the creation of prompts based on the image, and tagging of these prompts based on criteria such as reasoning, math reasoning, infographics, common knowledge questions, external knowledge questions, hyperspecific instruction following, chatbot, creative writing, and asking for advice. The model response was evaluated based on image understanding, instruction following, truthfulness, harmfulness, and verbosity.

This project involved image tagging (text free/light and text rich), as well as the creation of prompts based on the image, and tagging of these prompts based on criteria such as reasoning, math reasoning, infographics, common knowledge questions, external knowledge questions, hyperspecific instruction following, chatbot, creative writing, and asking for advice. The model response was evaluated based on image understanding, instruction following, truthfulness, harmfulness, and verbosity.

2024 - 2025

SFT project

Internal Proprietary ToolingImagePrompt Response Writing SFT

This project invovled generating prompts and responses, in an attempt to "break" the model by increasing the complexity of the the prompts in order to figure out the models limitations. Prompts were generated based on images, and the model was evaluated based on criteria such as truthfulness, relevance, accuracy, and conciseness.

This project invovled generating prompts and responses, in an attempt to "break" the model by increasing the complexity of the the prompts in order to figure out the models limitations. Prompts were generated based on images, and the model was evaluated based on criteria such as truthfulness, relevance, accuracy, and conciseness.

2024 - 2025

Education

Work History