Taya G.

Key Skills

Software

No software listed

Top Subject Matter

Agriculture (drone imagery for crop disease)

Conversational AI (sarcasm, intent, sentiment in customer support logs)

Security surveillance (anomaly detection, person re identification)

Top Data Types

Text

Top Task Types

Text Generation

Question Answering

Text Summarization

Freelancer Overview

ver the past several years, I have worked extensively on data labeling and AI training data across both computer vision and NLP projects. For vision tasks, I have handled polygon segmentation, bounding boxes, and keypoint annotation primarily for autonomous driving and retail inventory use cases. On the language side, I have done text classification, entity recognition, and sentiment labeling, along with adversarial example creation to test model robustness. Beyond just executing tasks, I developed a habit of identifying gaps in labeling guidelines and proposing clarifications before they led to quality drift. I have also collaborated on small teams to resolve ambiguous edge cases. For instance, we defined consistent rules for what counts as a near miss in driving clips or how to handle sarcasm in customer reviews. That experience taught me how to align with other annotators and help draft clearer instruction manuals from scratch. I am proficient with tools like Labelbox, SuperAnnotate, and Prodigy, and I have written basic Python scripts to calculate inter annotator agreement and flag low confidence labels. I believe strong training data is not just about volume but about precision, consistency, and understanding how each label will affect model behavior downstream. That is the standard I bring to every project.

Education

Work History