For employers

Hire this AI Trainer

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

LoginCreate account
Invite to Job
Ary Prasetyo Cahyono

Ary Prasetyo Cahyono

Multimodal AI Evaluator – Computer Vision, NLP & STEM

Indonesia flagSouth Tangerang, Indonesia
$12.00/hrEntry LevelRoboflowLabelimgLabel Studio

Key Skills

Software

RoboflowRoboflow
LabelImgLabelImg
Label StudioLabel Studio

Top Subject Matter

Technology – IoT, System Automation & Hardware Integration
Engineering & Sciences – Physics, Data Analysis & STEM
Artificial Intelligence – Computer Vision, NLP & Robotics

Top Data Types

AudioAudio
ImageImage
TextText

Top Task Types

Bounding Box
Transcription
Classification
Object Detection
Computer Programming Coding

Freelancer Overview

My experience in AI evaluation and data processing stems from developing advanced hardware-software integrations, most notably utilizing the OpenAI Whisper model for high-accuracy speech recognition. During this project, I was deeply involved in evaluating the model's transcription accuracy, testing edge cases, and analyzing AI outputs to ensure precise translation of vocal commands into mechanical actuations. This hands-on integration work required a critical eye for AI logic, performance validation, and identifying hallucinations or misinterpretations—core skills directly applicable to Reinforcement Learning from Human Feedback (RLHF) and technical data annotation. Complementing my AI integration experience is a rigorous analytical foundation in Instrumentation Physics, where I specialize in complex data processing, statistical error analysis, and hardware/software QA testing. As a native Indonesian speaker with professional English proficiency, I bring a unique intersection of linguistic nuance and deep STEM expertise. This allows me to effectively assess technical reasoning, evaluate code or API integration outputs, and provide high-quality, fact-checked human feedback to improve both language and technical AI models.

Entry LevelEnglishIndonesianArabic

Labeling Experience

Voice Command QA & Audio Transcription for Robotic Actuation (Whisper AI)

AudioTranscription
This project involved the integration and rigorous evaluation of the OpenAI Whisper model to actuate an assistive robotic hand. The primary scope was to perform Quality Assurance (QA) and validate the accuracy of speech-to-text transcriptions for hardware control. My specific tasks included testing various audio inputs, analyzing the AI's transcription accuracy, identifying edge-case misinterpretations, and optimizing the logical mapping between the AI's text output and the C/C++ microcontroller execution. Quality measures adhered to included maintaining near-zero hallucination rates for safety-critical robotic movements and ensuring low-latency processing for real-time physical actuation.

This project involved the integration and rigorous evaluation of the OpenAI Whisper model to actuate an assistive robotic hand. The primary scope was to perform Quality Assurance (QA) and validate the accuracy of speech-to-text transcriptions for hardware control. My specific tasks included testing various audio inputs, analyzing the AI's transcription accuracy, identifying edge-case misinterpretations, and optimizing the logical mapping between the AI's text output and the C/C++ microcontroller execution. Quality measures adhered to included maintaining near-zero hallucination rates for safety-critical robotic movements and ensuring low-latency processing for real-time physical actuation.

2026 - Present

YOLOv8 Object Detection for Automated Waste Sorting

ImageBounding Box
This 8-month R&D project focused on automating inorganic waste management using Computer Vision. The primary scope involved preparing an end-to-end custom dataset to train a YOLOv8 object detection model. My specific data labeling tasks included manually reviewing image datasets and meticulously drawing precise bounding boxes around various types of inorganic waste using Roboflow. The project size encompassed a custom dataset of over 1,000+ images. To ensure high quality and accuracy, I adhered to strict quality measures, including consistent class categorization, edge-case handling for overlapping objects, and iterative dataset refinement based on the model's mean Average Precision (mAP) evaluation during training runs on Google Colab.

This 8-month R&D project focused on automating inorganic waste management using Computer Vision. The primary scope involved preparing an end-to-end custom dataset to train a YOLOv8 object detection model. My specific data labeling tasks included manually reviewing image datasets and meticulously drawing precise bounding boxes around various types of inorganic waste using Roboflow. The project size encompassed a custom dataset of over 1,000+ images. To ensure high quality and accuracy, I adhered to strict quality measures, including consistent class categorization, edge-case handling for overlapping objects, and iterative dataset refinement based on the model's mean Average Precision (mAP) evaluation during training runs on Google Colab.

2025 - 2025

Education

U

Universitas Islam Syarif Hidayatullah Jakarta

Bachelor of Science, Applied Physics (Instrumentation)

Bachelor of Science
2021 - 2026

Work History

S

Smart Urban Farming

Electrical Division Member

South Tangerang
2024 - 2024
B

Balai Besar Lembaga Minyak dan Gas Bumi (LEMIGAS)

Project Management & IoT Intern

South Jakarta
2024 - 2024