Dylan M.

Key Skills

Software

Don't disclose

Top Subject Matter

LLM evaluation

Rlhf Domain Expertise

prompt engineering for Indonesian language models

Top Data Types

Video

Text

Image

Top Task Types

RLHF

Segmentation

Freelancer Overview

AI Trainer & LLM Evaluation Specialist (Independent Research). Brings 2+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include n8n, Docker, and LLM APIs. Education includes Bachelor of Informatics Engineering, Universitas Bahaudin Mudhary Madura (2025) and Scholarship Program (AI Engineer), Pijak x IBM SkillBuild (2025). AI-training focus includes data types such as Computer Code, Programming, and Video and labeling workflows including RLHF, Segmentation, and Evaluation.

Labeling Experience

Freelance Video Labeler & Data Annotator (Atlas Capture)

Don't discloseVideoSegmentation

Performed frame-level video annotation and object labeling to support computer vision model training pipelines. Applied spatial labeling using bounding boxes, polygons, and semantic segmentation masks across diverse video datasets with high precision and consistency. Ensured delivery quality by following annotation guidelines and maintaining low inter-rater disagreement on batches. • Labeled objects at frame level for video datasets. • Used bounding boxes, polygons, and segmentation masks. • Applied strict guideline-based quality assurance. • Exported annotations in JSON/CSV for downstream ML integration.

2025 - Present

AI Trainer & LLM Evaluation Specialist (Independent Research)

RLHF

Conducted RLHF-style preference ranking by evaluating and ranking LLM-generated responses based on factual accuracy, contextual relevance, logical coherence, and instruction-following quality. Built multi-turn prompt architectures to stress-test LLM behavior on edge cases involving linguistic ambiguity and Indonesian cultural context. Performed systematic scoring of AI outputs across grounding, hallucination risk, completeness, and safety alignment dimensions. • Simulated reinforcement learning workflows via human-feedback-style evaluation. • Designed evaluation criteria spanning factuality, coherence, and safety. • Supported model output quality assessment for dataset curation. • Automated benchmarking workflow logging for repeated evaluations.

2025 - Present

AI Content Evaluator & Linguistic Data Specialist (Independent)

Don't discloseText

Reviewed AI-generated Indonesian texts for factual alignment, grammatical accuracy, and logical coherence as part of AI content evaluation and training-data quality control. Transformed raw text corpora into structured datasets by managing data formatting and conversions across document and spreadsheet workflows. Rewrote and reformatted advanced academic documents in formal Indonesian with strict grammatical precision and syntax adherence. • Evaluated Indonesian text quality for training suitability. • Structured corpora into formatted datasets for AI use cases. • Performed rewriting and formatting for language quality. • Ensured outputs met grammar and coherence standards.

2024 - 2025

Education

Work History