Wilslas Chepkui

Key Skills

Software

AWS SageMaker

CVAT

Doccano

Labelbox

Label Studio

Roboflow

Scale AI

Supervisely

V7 Labs

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Document

Image

Medical Dicom

Text

Video

Top Label Types

Action Recognition

Bounding Box

Cuboid

Point Key Point

Polygon

Tracking

Freelancer Overview

I am a detail-oriented data professional with hands-on experience in data annotation, labeling, and AI response evaluation. My background includes working on remote projects where I assessed AI-generated responses for accuracy and coherence, provided structured feedback, and labeled large text datasets to ensure consistency and quality. I have a strong foundation in machine learning basics, including supervised and unsupervised learning, regression, and classification, which enables me to understand and improve AI model performance. My strengths lie in analytical thinking, high accuracy, and the ability to follow structured instructions, making me well-suited for roles focused on data quality and AI training data.

Labeling Experience

RLHF and SFT for Large Language Models

Scale AITextRLHFEvaluation Rating

Participating in Reinforcement Learning from Human Feedback (RLHF) by ranking model-generated responses based on honesty, harmlessness, and helpfulness. I also draft "Golden Responses" for Supervised Fine-Tuning (SFT).

Participating in Reinforcement Learning from Human Feedback (RLHF) by ranking model-generated responses based on honesty, harmlessness, and helpfulness. I also draft "Golden Responses" for Supervised Fine-Tuning (SFT).

2024

Autonomous Vehicle Perception

CVATVideoBounding BoxTracking

Responsible for frame-by-frame annotation of urban traffic footage. Tasks involve tracking pedestrians, vehicles, and cyclists using bounding boxes with high temporal consistency. I utilize interpolation tools to ensure smooth tracking across occlusions.

Responsible for frame-by-frame annotation of urban traffic footage. Tasks involve tracking pedestrians, vehicles, and cyclists using bounding boxes with high temporal consistency. I utilize interpolation tools to ensure smooth tracking across occlusions.

2023

High-Precision Semantic Segmentation for Retail

LabelboxImagePolygonSegmentation

Labeled thousands of consumer goods images using pixel-perfect polygon segmentation. This project aimed to train a visual search engine to identify specific clothing items and household products from user-uploaded photos.

Labeled thousands of consumer goods images using pixel-perfect polygon segmentation. This project aimed to train a visual search engine to identify specific clothing items and household products from user-uploaded photos.

2022 - 2022

Medical Entity Extraction (NER)

DoccanoDocumentEntity Ner ClassificationText Summarization

Annotated anonymized medical reports to identify and categorize entities such as "Diagnosis," "Medication," "Dosage," and "Patient History." This project supported the development of an automated patient summary tool.

Annotated anonymized medical reports to identify and categorize entities such as "Diagnosis," "Medication," "Dosage," and "Patient History." This project supported the development of an automated patient summary tool.

2021 - 2022

Education

Work History