Kelvin M.

Key Skills

Software

No software listed

Top Subject Matter

LLM alignment

STEM/coding QA

RLHF and SFT training

Top Data Types

Text

Image

Top Task Types

RLHF

Segmentation

Freelancer Overview

AI Trainer, Outlier (LLM evaluation, prompt writing, preference data collection, and coding QA). Core strengths include Outlier and Remotask. Education includes Master of Business Administration, University of Nairobi (2020) and Bachelor of Human Resource Management, Moi University (2010). AI-training focus includes data types such as Text and Image and labeling workflows including RLHF and Segmentation.

Labeling Experience

AI Trainer — Outlier (2022 – Present)

TextRLHF

Evaluated and ranked LLM responses for helpfulness, harmlessness, and truthfulness across STEM, coding, and general knowledge domains. Authored and refined prompts for supervised fine-tuning to improve model reasoning and instruction following. Collected preference data for RLHF by comparing model outputs using detailed rubrics. • Ranked outputs using evaluation criteria (helpfulness, harmlessness, truthfulness). • Wrote/refined prompts for SFT-style instruction following. • Performed preference data collection for RLHF training. • Conducted code review and debugging to support coding-capable LLMs in Python, SQL, and JavaScript.

2022 - Present

Data Labeling Specialist — Remotask (2020 – 2022)

ImageSegmentation

Labeled images and video frames to support autonomous driving and retail analytics AI models. Performed 2D/3D bounding box annotation, semantic segmentation, and keypoint labeling with 98%+ accuracy. Completed LiDAR point cloud annotation tasks for 3D object detection and sensor fusion. • Annotated 2D/3D bounding boxes for object localization. • Labeled semantic segmentation and keypoints for scene understanding. • Annotated LiDAR point clouds for 3D object detection and sensor fusion. • Reviewed peer submissions, provided feedback, and supported onboarding new labelers to improve quality.

2020 - 2022

Education