Keshav Agarwal

Key Skills

Software

AWS SageMaker

Google Cloud Vertex AI

Other

Top Subject Matter

AI-FAQ

Multilingual LLMs

Text processing

Top Data Types

Document

Text

Top Task Types

Classification

Fine-tuning

Text Generation

Text Summarization

Translation/Localization

Freelancer Overview

As a skilled AI enthusiast with a solid foundation in data labeling and AI training data, I have honed my expertise through both academic pursuits and practical experiences. My technical proficiencies include Python, C++, and JavaScript, alongside advanced machine learning frameworks such as PyTorch, TensorFlow, and QLora. I have actively contributed to projects involving Convolutional Neural Networks (CNNs), YOLO, Transformers, and Generative Adversarial Networks (GANs). My experience is further augmented by hands-on work with tools like Docker, AWS, and Git. One of my notable projects, "GITA GPT," involved developing an AI chatbot leveraging a fine-tuned 7 billion parameter Llama 2 model for interactive user queries. I utilized AWS SageMaker for deployment, achieving significant improvements in response accuracy and relevance. Another significant project, "ClinicalSentix," showcased my ability to apply advanced NLP techniques and data visualization to create a robust sentiment analysis tool. Additionally, in the "NeuraC" project, I developed a neural network framework and real-time visualization tool using C and Raylib, optimizing model training and ensuring cross-platform compatibility. These projects highlight my ability to integrate cutting-edge technologies and deliver impactful AI solutions, setting me apart as a proficient contributor in the AI and data labeling domain.

Labeling Experience

Clinical Sentix

OtherTextEntity Ner ClassificationRelationship

Applied Advanced NLP techniques, fine-tuned RoBERTa model achieving 90% sentiment analysis accuracy, and performed topic modeling to enhance data interpretability. • Built a tool similar to Brand Watch with robust backend with FastAPI and MongoDB, efficiently managing and preprocessing extensive datasets from Twitter, Reddit, and Drug.com. • Developed an interactive React and Tremor dashboard for visualizing sentiment distribution, topic modeling, and time series analysis, enabling comparative drug analysis and improved decision-making.

Applied Advanced NLP techniques, fine-tuned RoBERTa model achieving 90% sentiment analysis accuracy, and performed topic modeling to enhance data interpretability. • Built a tool similar to Brand Watch with robust backend with FastAPI and MongoDB, efficiently managing and preprocessing extensive datasets from Twitter, Reddit, and Drug.com. • Developed an interactive React and Tremor dashboard for visualizing sentiment distribution, topic modeling, and time series analysis, enabling comparative drug analysis and improved decision-making.

2024

GITA GPT

Aws SagemakerTextQuestion AnsweringText Generation

Developed an AI chatbot using Three.js and a text-to-speech API, incorporating a fine-tuned 7 billion parameter Llama 2 model specifically trained on the Bhagavad Gita dataset for engaging interactions. • Utilized AWS SageMaker to deploy 7 billion parameter Llama 2, achieving a 30% increase in query response accuracy. Leveraging its API for user query responses through the chatbot to the Custom Model. • Implemented dual-model query processing, integrating custom and GPT-4 models. Leading to 25% improvement in response relevance. Integrated OpenAI Whisper for TTS synthesis for clear and accurate voice output.

Developed an AI chatbot using Three.js and a text-to-speech API, incorporating a fine-tuned 7 billion parameter Llama 2 model specifically trained on the Bhagavad Gita dataset for engaging interactions. • Utilized AWS SageMaker to deploy 7 billion parameter Llama 2, achieving a 30% increase in query response accuracy. Leveraging its API for user query responses through the chatbot to the Custom Model. • Implemented dual-model query processing, integrating custom and GPT-4 models. Leading to 25% improvement in response relevance. Integrated OpenAI Whisper for TTS synthesis for clear and accurate voice output.

2024 - 2024

Education

Work History