Mariam Gogichaishvili

Key Skills

Software

CVAT

Top Subject Matter

No subject matter listed

Top Data Types

Image

Top Label Types

Bounding Box

Classification

Freelancer Overview

Detail-oriented Visual Data Annotator and Professional Interpreter with a Bachelor’s in Visual Arts and Design and recent certification in AI Essentials (2025). I specialize in the high-precision labeling of visual datasets and the validation of multilingual AI responses. With 8+ years of experience in translation and linguistic analysis, I bring a unique 'multimodal' perspective to AI training—ensuring that complex visual data is accurately categorized and contextually described across English, Georgian, Russian, and French. My background as a visual artist allows me to execute technical annotation tasks (2D/3D bounding boxes, polygon segmentation, and key-point landmarking) with pixel-perfect accuracy. Simultaneously, my linguistic expertise enables me to evaluate 'Text-to-Image' and 'Image-to-Text' models for cultural nuance and grammatical precision. I am proficient in CVAT and Labelbox, and I am committed to delivering the high-quality human-in-the-loop feedback necessary for the next generation of global AI models.

Labeling Experience

Visual Data Annotation & Image Segmentation for AI Training

CVATImageBounding BoxClassification

Using the open-source CVAT platform, I performed pixel-level labeling on a dataset of complex street scenes. I focused on 'occluded objects' (e.g., a pedestrian partially hidden by a car) and 'edge-case' environmental conditions like rain and glare. My process involved tracing intricate polygons around 20+ classes including sidewalks, lane markings, and traffic signals, maintaining a 99% boundary accuracy rate. This project demonstrated my ability to follow strict annotation protocols and deliver model-ready ground truth data.

Using the open-source CVAT platform, I performed pixel-level labeling on a dataset of complex street scenes. I focused on 'occluded objects' (e.g., a pedestrian partially hidden by a car) and 'edge-case' environmental conditions like rain and glare. My process involved tracing intricate polygons around 20+ classes including sidewalks, lane markings, and traffic signals, maintaining a 99% boundary accuracy rate. This project demonstrated my ability to follow strict annotation protocols and deliver model-ready ground truth data.

2025 - 2025

Education

Work History