For employers

Hire this AI Trainer

Sign in or create an account to invite AI Trainers to your job.

Invite to Job
Edwin Stevanus

Edwin Stevanus

Visual Prompt Engineer | Computer Vision & Annotation Specialist

INDONESIA flag
Bandung, Indonesia
$17.50/hrIntermediateCVATAppenClickworker

Key Skills

Software

CVATCVAT
AppenAppen
ClickworkerClickworker
Internal/Proprietary Tooling
MindriftMindrift
Scale AIScale AI
TolokaToloka
RoboflowRoboflow

Top Subject Matter

Computer Vision
Multimodal AI
Visual Logic Annotation

Top Data Types

VideoVideo
TextText
ImageImage

Top Task Types

Segmentation
Red Teaming
Bounding Box
Question Answering
Text Summarization
RLHF
Object Detection
Point Key Point
Classification
Polygon

Freelancer Overview

I am a Visual Logic Architect with intermediate-level AI training experience, built upon nearly two decades of professional spatial design and QA formatting. My core expertise lies in Computer Vision (CV) annotation, specifically high-precision bounding box conceptualization, frame-by-frame temporal logic, and visual occlusion mapping for motion analytics. I specialize in translating complex real-world geometry into pixel-perfect training data using specialized labeling platforms. Beyond visual segmentation, I am highly experienced in LLM Red Teaming, adversarial logic testing, and cross-lingual semantic evaluation (Native Indonesian / Fluent English). I excel at identifying systemic vulnerabilities in AI protocols and enforcing strict technical constraints across both visual and multimodal generative models.

IntermediateEnglishIndonesianSundanese

Labeling Experience

Visual Prompt Engineer & AI Production Director

OtherVideoSegmentation
Designed and executed continuous-take generative AI video sequences with strict structural logic, crafting precise AI training data. Utilized spatial and temporal logic to ensure pixel-perfect segmentation and frame-perfect multimodal synchronization of visual and audio data. Programmed technical chassis, lighting, and multi-character logic for high-fidelity video annotation workflows. • Directed frame-by-frame segmentation and synchronization using AI-assisted video editing tools. • Enforced rigid visual formatting and aspect ratio constraints on training data assets. • Labeled temporal transitions, character movements, and SFX correlations for multimodal datasets. • Developed internal protocols for consistency in generative output and stage logic.

Designed and executed continuous-take generative AI video sequences with strict structural logic, crafting precise AI training data. Utilized spatial and temporal logic to ensure pixel-perfect segmentation and frame-perfect multimodal synchronization of visual and audio data. Programmed technical chassis, lighting, and multi-character logic for high-fidelity video annotation workflows. • Directed frame-by-frame segmentation and synchronization using AI-assisted video editing tools. • Enforced rigid visual formatting and aspect ratio constraints on training data assets. • Labeled temporal transitions, character movements, and SFX correlations for multimodal datasets. • Developed internal protocols for consistency in generative output and stage logic.

2024 - Present

Adversarial AI Logic Tester (Independent Research)

OtherTextRed Teaming
Conducted adversarial, semantic, and logical boundary testing on LLM-powered financial AI systems for robustness evaluation. Developed and introduced edge-case scenarios to identify vulnerabilities and logical inconsistencies in automated protocols. Documented and labeled semantic contradictions and failures for systematic model improvement. • Simulated adversarial prompts and evaluated protocol responses in financial AI. • Designed and tested prompts for instruction drift and logic loop analysis. • Identified points of semantic confusion and guardrail bypasses. • Assigned and categorized example outcomes for model red-teaming datasets.

Conducted adversarial, semantic, and logical boundary testing on LLM-powered financial AI systems for robustness evaluation. Developed and introduced edge-case scenarios to identify vulnerabilities and logical inconsistencies in automated protocols. Documented and labeled semantic contradictions and failures for systematic model improvement. • Simulated adversarial prompts and evaluated protocol responses in financial AI. • Designed and tested prompts for instruction drift and logic loop analysis. • Identified points of semantic confusion and guardrail bypasses. • Assigned and categorized example outcomes for model red-teaming datasets.

2024 - 2024

Education

U

Universitas Kristen Maranatha

Bachelor of Arts, Visual Communication Design

Bachelor of Arts
2005 - 2009

Work History

H

Hybrid AI

Community AI Safety, Chat & Content Moderator

Bandung
2024 - Present
F

Freelance

Senior Graphic Designer & Brand Architect

Bandung
2005 - Present