Kibae K.

Key Skills

Software

No software listed

Top Subject Matter

AI Training

Mathematics

Finance

Top Data Types

Image

Text

Document

Top Task Types

Data Collection

Computer Programming/Coding

Evaluation/Rating

RLHF

Classification

Freelancer Overview

Most of my current work at Handshake AI is evaluation focused. I look at AI-generated images and assess them across a few dimensions, whether the output actually reflects the prompt, how realistic it is, and what specifically broke down when it did not deliver. A chunk of my role also involves writing the test prompts we use, and I tend to gravitate toward the harder ones, scenarios with built-in ambiguity or conditions that are likely to expose gaps in how the model generalizes. Peer review is part of it too. I check other evaluators' submissions regularly to make sure scoring stays consistent. Years of quantitative work trained me to look for where logic breaks, and that transfers directly.

Labeling Experience

AI Trainer

ImageRLHF

As an AI Trainer at Handshake AI, I evaluated AI-generated images using structured quality rubrics. My responsibilities included scoring visual accuracy, realism, and prompt adherence in generated outputs. I also conducted edge-case failure analysis to support systematic model improvement. • Designed and refined prompt-based evaluation scenarios and scoring frameworks • Reviewed submissions from peers to ensure evaluation consistency and standard adherence • Supported cross-team calibration via rubric alignment and structured feedback • Utilized internal/proprietary tools specific to Handshake AI for labeling and assessment

2025 - Present

Education

Work History