Deva D.

Key Skills

Software

Don't disclose

Top Subject Matter

Politics

Education

artefacts

Top Data Types

Image

Text

Document

Top Task Types

Polyline

Entity (NER) Classification

Cuboid

Freelancer Overview

I am a PhD researcher at the University of Bristol and a currently active AI Evaluator with Mercor, where I assess AI-generated responses against detailed project rubrics, evaluating coherence, reasoning quality, factual accuracy, and nuance. I have completed qualification assessments for OpenAI's Les Artistes – Artifacts project, including Screener, Quiz, and High-Level Grading tasks, and consistently produce all rationales through independent critical analysis, entirely without the use of AI writing tools. My academic background in critical discourse analysis and qualitative research has directly equipped me to identify implicit reasoning, logical gaps, and subtle errors in complex textual outputs at scale. What sets me apart is the depth of evaluative judgment I bring from years of high-stakes academic assessment. As a Graduate Teaching Assistant at Bristol, I designed marking frameworks and delivered structured formative feedback on student work, skills that translate directly into consistent, rubric-aligned AI output evaluation. My PhD research involved building a 192-reference NVivo coding architecture and conducting cross-lingual corpus analysis across Tamil, Sinhala, and English sources, demonstrating my ability to handle diverse, complex data with rigour and precision. Combined with native fluency in English and Tamil, and professional proficiency in Hindi, Telugu, and Malayalam, I bring both analytical depth and multilingual range to AI training and data labeling projects.

Labeling Experience

Mercor

TextEvaluation/Rating

Contributed to an AI training project in collaboration with OpenAI, focused on evaluating and improving the quality of AI-generated creative and artifact-based outputs. Completed a multi-stage qualification process comprising the Artifacts Trainer Screener, Artifacts Quiz, and High-Level (HL) Grading tasks, demonstrating proficiency in applying structured evaluation criteria to complex AI outputs. Core responsibilities involved assessing AI responses for reasoning quality, coherence, factual accuracy, and nuance using detailed project rubrics, and producing precise written rationales to support model improvement. All evaluations were conducted independently and without the use of AI writing tools.

2026 - Present

Education

Work History