Maneesha Rajaratne

Key Skills

Software

Data Annotation Tech

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Computer Code Programming

Document

Top Label Types

Mapping

Point Key Point

Prompt Response Writing SFT

Freelancer Overview

I am an experienced data analyst and annotator with a strong background in AI training data, data labeling, and data analytics. My expertise includes hands-on annotation and training of large language models (LLMs), where I have delivered high-quality annotated datasets and optimized workflows for model development. I am proficient in Python, data visualization, and machine learning techniques, and have worked extensively with tools such as Azure Data Factory, Tableau, and SQL. My work spans domains including financial fraud detection, healthcare (e-Health applications and breast cancer prediction), and natural language processing. I am skilled at identifying data quality issues, developing ETL pipelines, and communicating complex data findings to both technical and non-technical stakeholders, ensuring the highest standards of data integrity and usability for AI and analytics projects.

Labeling Experience

Rate the personality calibration demonstrated in an audio response

Data Annotation TechAudioPoint Key Point

The scope of the project is to identify if the audio response the model generated contains the assigned rubric and followed the instructions well based on the prompt explicit and implicit instructions while preserving the assigned rubric. For an example if the given prompt is assigned as the rubric 'Humor', the response must contains appropriate level of humor in order to rate the response as perfect. If the response doesn't contain humor or failed to understand the prompt and responded to the literal meaning of the prompt, response is rated as under-utilized and if the response has too much humor that doesn't really focus the context for the prompt, it is rated as over-utilized. Like wise there are several rubrics to check like, the Insightful, Straightforwardness, Authenticity, Empathy etc. This project purpose it solely to make the AI understand different aspects of the human responses, and make it less robotic when responding to different situations.

The scope of the project is to identify if the audio response the model generated contains the assigned rubric and followed the instructions well based on the prompt explicit and implicit instructions while preserving the assigned rubric. For an example if the given prompt is assigned as the rubric 'Humor', the response must contains appropriate level of humor in order to rate the response as perfect. If the response doesn't contain humor or failed to understand the prompt and responded to the literal meaning of the prompt, response is rated as under-utilized and if the response has too much humor that doesn't really focus the context for the prompt, it is rated as over-utilized. Like wise there are several rubrics to check like, the Insightful, Straightforwardness, Authenticity, Empathy etc. This project purpose it solely to make the AI understand different aspects of the human responses, and make it less robotic when responding to different situations.

2025

Education

Work History