Vaibhav Shastri

Key Skills

Software

Appen

Clickworker

Img Lab

Scale AI

Data Annotation Tech

Top Subject Matter

No subject matter listed

Top Data Types

Document

Image

Text

Top Label Types

Mapping

Segmentation

Text Summarization

Freelancer Overview

I bring hands-on experience in data labeling and AI training through my work on the Outlier Project, where I designed and evaluated complex image-based prompts to test AI models’ reasoning and bias-handling capabilities. This involved using tools like ChatGPT 4.0, Gemini Advanced, WolframAlpha, and platforms like Welocalize and DataAnnotation to refine model outputs and identify knowledge gaps. My contribution directly supported enhancing the accuracy and performance of AI systems, with a focus on fairness, contextual understanding, and prompt optimization. My academic foundation in Computer Science Engineering and MBA in Marketing & Analytics, combined with strong exposure to data-centric tools like SPSS, PowerBI, Excel (PowerPivot, Macros, PowerQuery) and survey platforms, has equipped me to handle structured and unstructured data effectively. This unique blend of technical and analytical skills enables me to contribute meaningfully to AI training initiatives, especially where nuanced prompt creation, data validation, or behavior testing of LLMs is critical.

Labeling Experience

Outlier AI Prompt Evaluation & Annotation Project

Data Annotation TechImageText GenerationEmotion Recognition

Worked on a high-impact annotation project focused on enhancing the reasoning and ethical performance of AI models. The scope included creating image-based and text-based prompts designed to assess model capabilities across problem-solving, ethical reasoning, and bias detection. Tasks involved annotating model responses, labeling logical inconsistencies, identifying hallucinations, and flagging harmful or inappropriate content. Delivered detailed feedback loops to guide model improvement, involving multiple annotation types like emotion tagging, named entity recognition, and red-teaming edge-case prompts. The project helped shape model training strategies by contributing over 500+ labeled samples, with adherence to strict quality guidelines, benchmark rubrics, and weekly performance audits.

Worked on a high-impact annotation project focused on enhancing the reasoning and ethical performance of AI models. The scope included creating image-based and text-based prompts designed to assess model capabilities across problem-solving, ethical reasoning, and bias detection. Tasks involved annotating model responses, labeling logical inconsistencies, identifying hallucinations, and flagging harmful or inappropriate content. Delivered detailed feedback loops to guide model improvement, involving multiple annotation types like emotion tagging, named entity recognition, and red-teaming edge-case prompts. The project helped shape model training strategies by contributing over 500+ labeled samples, with adherence to strict quality guidelines, benchmark rubrics, and weekly performance audits.

2024 - 2024

Education

Work History