Madhavi Latha

Key Skills

Software

Appen

CrowdSource

CVAT

Data Annotation Tech

Labelbox

OneForma

Remotasks

Scale AI

Toloka

Telus

Other

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Image

Text

Top Label Types

Audio Recording

Bounding Box

Computer Programming Coding

Entity Ner Classification

Translation Localization

Freelancer Overview

I am an AI Training Expert with extensive experience in data labeling, model evaluation, and prompt development across text, image, and audio data. My work spans LLM evaluation, content moderation, and multilingual AI training in English, Hindi, French, German, and Spanish. I have successfully delivered projects involving response accuracy checks, content quality assurance, sentiment analysis, named entity recognition, and classification tasks, ensuring outputs meet the highest standards of compliance, clarity, and contextual relevance. Alongside AI training, I bring over 6 years of professional expertise in operations, consulting, and financial analysis, which allows me to apply domain knowledge in business, insurance, and manufacturing to AI projects. I am proficient with leading data labeling tools and platforms (Labelbox, CVAT,Scale AI etc) and skilled in ensuring multimodal annotation for text, image, video, and voice datasets. With a proven track record of enhancing AI systems through quality evaluation, precise annotation, and guideline-based improvements, I am committed to advancing cutting-edge AI models across diverse industries.

Labeling Experience

Evaluation & Rating of AI-Generated Responses

OtherDocumentRLHFFine Tuning

Rated AI-generated responses on an 8-point scale for accuracy, fluency, completeness, and formatting. Annotated datasets to fine-tune models for human-like interaction. Identified issues such as factual inaccuracies, grammatical errors, and unsuitable tone. Provided rewrites and improvements, ensuring models generate clear, contextual, and high-quality responses. Supported fine-tuning pipelines by curating datasets for better reasoning and dialogue flow.

Rated AI-generated responses on an 8-point scale for accuracy, fluency, completeness, and formatting. Annotated datasets to fine-tune models for human-like interaction. Identified issues such as factual inaccuracies, grammatical errors, and unsuitable tone. Provided rewrites and improvements, ensuring models generate clear, contextual, and high-quality responses. Supported fine-tuning pipelines by curating datasets for better reasoning and dialogue flow.

2024

voice and video analysis

OneformaAudioText GenerationEmotion Recognition

I have done multi-lingual data annotation, transcription, classification, QA in English, Hindi, and other languages.

I have done multi-lingual data annotation, transcription, classification, QA in English, Hindi, and other languages.

2023

Image & Voice Data Annotation for Multimodal AI

Scale AIImageEntity Ner ClassificationClassification

Labeled image-based responses for relevance, compliance, and accuracy. Annotated voice datasets for clarity, tone, intonation, and instruction-following. Classified user sentiment and emotional tone in spoken responses. Performed segmentation tasks on visual datasets to improve multimodal AI alignment. Ensured cross-modal consistency between text, image, and audio responses.

Labeled image-based responses for relevance, compliance, and accuracy. Annotated voice datasets for clarity, tone, intonation, and instruction-following. Classified user sentiment and emotional tone in spoken responses. Performed segmentation tasks on visual datasets to improve multimodal AI alignment. Ensured cross-modal consistency between text, image, and audio responses.

2023

AI Content Moderation, Prompt Writing & Annotation

Scale AITextText GenerationRLHF

Reviewed and annotated text, image, and multimodal AI outputs to ensure compliance with safety guidelines. Classified data into safe, unsafe, ambiguous, or non-compliant categories. Performed red teaming to stress-test AI systems for bias, toxicity, and unsafe responses. Annotated relationship-based classifications to differentiate intent, tone, and user sensitivity. Applied privacy and compliance standards to protect user data and ensure responsible AI. Annotated plugin-related prompts to train AI models in correctly identifying triggers for real-time vs. knowledge-based responses. Classified prompts by intent, domain relevance, and execution type (function calling vs. general reasoning). Performed relationship labeling to map queries with suitable actions. Labeled Q&A datasets for clarity, reasoning accuracy, and contextual alignment. Ensured consistency and reliability in plugin-based AI deployments.

Reviewed and annotated text, image, and multimodal AI outputs to ensure compliance with safety guidelines. Classified data into safe, unsafe, ambiguous, or non-compliant categories. Performed red teaming to stress-test AI systems for bias, toxicity, and unsafe responses. Annotated relationship-based classifications to differentiate intent, tone, and user sensitivity. Applied privacy and compliance standards to protect user data and ensure responsible AI. Annotated plugin-related prompts to train AI models in correctly identifying triggers for real-time vs. knowledge-based responses. Classified prompts by intent, domain relevance, and execution type (function calling vs. general reasoning). Performed relationship labeling to map queries with suitable actions. Labeled Q&A datasets for clarity, reasoning accuracy, and contextual alignment. Ensured consistency and reliability in plugin-based AI deployments.

2023

Text Categorization & Classification for AI Model Training

AppenTextEntity Ner ClassificationClassification

Annotated large volumes of AI-generated text responses to improve accuracy, fluency, sentiment, tone, and contextual relevance. Categorized prompts into types such as instruction-following, creative writing, factual Q&A, and gibberish detection. Labeled entities (NER) and provided text summarization to train LLMs on concise, human-like output. Ensured multilingual labeling (English, Hindi, Telugu, French, German, Spanish) for localization and cultural accuracy. Maintained high-quality standards through detailed evaluation and justification reports.

Annotated large volumes of AI-generated text responses to improve accuracy, fluency, sentiment, tone, and contextual relevance. Categorized prompts into types such as instruction-following, creative writing, factual Q&A, and gibberish detection. Labeled entities (NER) and provided text summarization to train LLMs on concise, human-like output. Ensured multilingual labeling (English, Hindi, Telugu, French, German, Spanish) for localization and cultural accuracy. Maintained high-quality standards through detailed evaluation and justification reports.

2023

Education

Work History