Lekya Sree

Key Skills

Software

AWS SageMaker

Top Subject Matter

No subject matter listed

Top Data Types

Text

Top Label Types

Segmentation

Classification

Question Answering

Text Generation

Text Summarization

Freelancer Overview

I am a Machine Learning Data Associate Engineer with hands-on experience in developing and maintaining production-grade AI data workflows, including LLM pipelines, RAG systems, and prompt evaluation frameworks for high-volume enterprise applications. My expertise spans data labeling, annotation, and validation, where I have implemented robust pipelines for relevance scoring, correctness checks, and consistency validation—directly contributing to improved model performance and reduced downstream errors. I am skilled in Python, TensorFlow, PyTorch, and Scikit-learn, and have automated quality checks and monitoring to ensure high reliability and productivity in AI training data processes. My background also includes deploying ML models for NLP and audio classification projects, collaborating in Agile teams, and consistently delivering high-quality, scalable solutions in fast-paced environments.

Labeling Experience

AI Model Output Evaluation & Response Classification (RAG Systems)

Aws SagemakerTextSegmentationClassification

Worked on AI-driven RAG and Non-RAG workflows involving structured evaluation and classification of model-generated responses. Assessed outputs for accuracy, relevance, factual consistency, and adherence to annotation guidelines. Performed comparative evaluation and quality rating of responses to support dataset refinement and model improvement. Maintained 96%+ quality reliability while exceeding productivity targets, ensuring consistent labeling standards and reduced rework. Contributed to identifying edge cases and improving guideline clarity for better annotation consistency.

Worked on AI-driven RAG and Non-RAG workflows involving structured evaluation and classification of model-generated responses. Assessed outputs for accuracy, relevance, factual consistency, and adherence to annotation guidelines. Performed comparative evaluation and quality rating of responses to support dataset refinement and model improvement. Maintained 96%+ quality reliability while exceeding productivity targets, ensuring consistent labeling standards and reduced rework. Contributed to identifying edge cases and improving guideline clarity for better annotation consistency.

2024 - 2025

Education

Work History