Anish Mandal

Key Skills

Software

Other

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Document

Image

Text

Top Label Types

Action Recognition

Audio Recording

Classification

Diagnosis

Evaluation Rating

Fine Tuning

Prompt Response Writing SFT

Question Answering

RLHF

Segmentation

Text Generation

Text Summarization

Translation Localization

Freelancer Overview

I am an experienced data annotator and AI model trainer with a strong background in data labeling, prompt engineering, and quality assurance for AI systems. Over the past two years, I’ve transitioned into AI-focused work, like on "Outlier AI (Scale AI)," "Soul AI (Decan AI)," and "Turing AI," where my role involved creating complex prompts, identifying reasoning failures, correcting chain-of-thought steps, and guiding models to verified ground-truth answers, mainly in STEM and logic-heavy tasks. I also worked on evaluation and comparison projects, where I assessed model responses using detailed rubrics - including instruction following, factual accuracy, reasoning quality, relevance, tone, and formatting - across different types of data, including text, audio, image annotation. I am skilled in using tools such as Power BI, Python, and Generative AI platforms to analyze data and streamline processes. My projects have spanned domains like natural language processing, e-commerce, and business analytics, and I have consistently maintained high-quality standards and top-rated performance in AI training tasks. I am fluent in English, Bengali, and Hindi, and bring a detail-oriented, analytical approach to every data labeling project I undertake.

Labeling Experience

Business Analyst

OtherTextQuestion AnsweringText Generation

I have been working on "Google Multimodal GDM" project for "Turing AI" on "Populous Studio," evaluating responses on different parameters, like 'Task Completion,' 'Factual Accuracy,' 'Relevance,' 'Tone,' 'Formatting,' etc. of the responses to user prompts, based on pre-defined Rubric Guidelines.

I have been working on "Google Multimodal GDM" project for "Turing AI" on "Populous Studio," evaluating responses on different parameters, like 'Task Completion,' 'Factual Accuracy,' 'Relevance,' 'Tone,' 'Formatting,' etc. of the responses to user prompts, based on pre-defined Rubric Guidelines.

2025

AI Model Trainer & Quality Analyst

Scale AITextTranslation LocalizationRLHF

• As an AI Model Trainer & Quality Analyst at "Outlier AI" (daughter company of "Scale AI," I annotated and reviewed over 150 tasks for training and consistency. I ensured outputs adhered to guidelines and proper ground truth answers. My responsibilities supported platform quality and satisfaction standards. • Annotated and refined prompts/tasks for accuracy and adherence. • Reviewed AI model responses against ground truth final answers. • Streamlined QA protocols and performed guideline checks. • Achieved a top-rated status with 100% client satisfaction.

• As an AI Model Trainer & Quality Analyst at "Outlier AI" (daughter company of "Scale AI," I annotated and reviewed over 150 tasks for training and consistency. I ensured outputs adhered to guidelines and proper ground truth answers. My responsibilities supported platform quality and satisfaction standards. • Annotated and refined prompts/tasks for accuracy and adherence. • Reviewed AI model responses against ground truth final answers. • Streamlined QA protocols and performed guideline checks. • Achieved a top-rated status with 100% client satisfaction.

2025 - 2025

Data Annotator & Evaluator

OtherTextSegmentationRLHF

• As a Data Annotator at Soul AI, I evaluated the factual accuracy and objectivity of AI model responses across diverse prompts. I refined unsupported claims and contributed to improved AI reliability. My insights helped enhance response structures and user satisfaction. • Assessed truthfulness, objectivity, and subjectivity in AI-generated text outputs. • Identified and corrected over 45 disputed claims within prompts. • Provided structured feedback for hedging and framing improvements. • Contributed to a 15% increase in user satisfaction by refining results.

• As a Data Annotator at Soul AI, I evaluated the factual accuracy and objectivity of AI model responses across diverse prompts. I refined unsupported claims and contributed to improved AI reliability. My insights helped enhance response structures and user satisfaction. • Assessed truthfulness, objectivity, and subjectivity in AI-generated text outputs. • Identified and corrected over 45 disputed claims within prompts. • Provided structured feedback for hedging and framing improvements. • Contributed to a 15% increase in user satisfaction by refining results.

2024 - 2025

Education

Work History