Arman Hossain

Key Skills

Software

Appen

Data Annotation Tech

Telus

Top Subject Matter

No subject matter listed

Top Data Types

Document

Image

Video

Top Label Types

Emotion Recognition

Evaluation Rating

Mapping

Object Detection

Segmentation

Freelancer Overview

I have over a year of hands-on experience in AI training and data labeling, working on diverse projects spanning social media, gaming, and multimedia content. My work has included mapping complex maze games for LLM training, extracting and annotating entities from social media videos and images, and performing reverse prompt engineering to guide AI in generating accurate visual and textual outputs. Due to the quality and impact of my work, I was promoted and entrusted with auditing and reviewing the work of others, ensuring high standards across all projects. In addition to labeling and annotation, I specialize in training AI for real-time recognition of humans, objects, and scenes, helping models accurately interpret dynamic content. I combine technical expertise with attention to detail and creative problem-solving, producing structured, high-quality datasets that enable AI to perform reliably across varied and complex scenarios.

Labeling Experience

Instagram image entity tagging

TelusImageSegmentationClassification

Analyzed Instagram images to identify and tag all visible entities, including people, clothing, style, animals, locations, and more. Each entity was paired with alternate reference images to provide rich context. This structured labeling helps AI models understand visual content, recognize objects and people accurately, and improve image-based recommendations, content moderation, and visual search applications.

Analyzed Instagram images to identify and tag all visible entities, including people, clothing, style, animals, locations, and more. Each entity was paired with alternate reference images to provide rich context. This structured labeling helps AI models understand visual content, recognize objects and people accurately, and improve image-based recommendations, content moderation, and visual search applications.

2025 - 2025

AI Video Prompt Reverse Engineering

TelusVideoClassificationEmotion Recognition

Examined Instagram videos to deconstruct AI-generated or inspired content, identifying key visual and temporal elements, scene compositions, actions, and styles to infer underlying prompts. This enabled AI to generate or recreate videos with greater accuracy and fidelity. Supports AI in understanding temporal dynamics and multi-frame content generation, enhancing applications in video synthesis, creative media, motion tracking, real-time animation, and content personalization.

Examined Instagram videos to deconstruct AI-generated or inspired content, identifying key visual and temporal elements, scene compositions, actions, and styles to infer underlying prompts. This enabled AI to generate or recreate videos with greater accuracy and fidelity. Supports AI in understanding temporal dynamics and multi-frame content generation, enhancing applications in video synthesis, creative media, motion tracking, real-time animation, and content personalization.

2025

AI Image Prompt Reverse Engineering

TelusTextClassificationText Generation

Analyzed Instagram images to reverse-engineer prompts used in AI-generated content, identifying visual elements, styles, compositions, and attributes that influenced generation. This enabled precise reconstruction of prompts to guide AI for consistent and high-quality outputs. Helps AI models learn correlations between textual prompts and visual outputs, improving image generation, content creation tools, design assistance, and visual AI research.

Analyzed Instagram images to reverse-engineer prompts used in AI-generated content, identifying visual elements, styles, compositions, and attributes that influenced generation. This enabled precise reconstruction of prompts to guide AI for consistent and high-quality outputs. Helps AI models learn correlations between textual prompts and visual outputs, improving image generation, content creation tools, design assistance, and visual AI research.

2025

Processed Instagram videos to detect and tag all entities frame by frame, including people, attire, style, animals, locations, and actions. Entities were cross-referenced with alternate images to enhance AI understanding. This enables AI to interpret dynamic content, track objects and people over time, and improve applications in video summarization, real-time recognition, and content analysis.

TelusVideoClassificationObject Detection

Processed Instagram videos to detect and tag all entities frame by frame, including people, attire, style, animals, locations, and actions. Entities were cross-referenced with alternate images to enhance AI understanding. This enables AI to interpret dynamic content, track objects and people over time, and improve applications in video summarization, real-time recognition, and content analysis.

Processed Instagram videos to detect and tag all entities frame by frame, including people, attire, style, animals, locations, and actions. Entities were cross-referenced with alternate images to enhance AI understanding. This enables AI to interpret dynamic content, track objects and people over time, and improve applications in video summarization, real-time recognition, and content analysis.

2025

Education

Work History