ML Engineer | Ex-ML Research & Development @IIT-H (Vigil Labs) | Data Scientist | MLOps | Production Systems

medium

leetcode

github

projects

Divyansh Pandey  • Peerlist

Developed an interactive data analysis toolkit designed to streamline and enhance various data analysis tasks, from exploratory data analysis to advanced feature engineering and model evaluation. This project provides a comprehensive suite of tools for data scientists, analysts, and enthusiasts to efficiently clean, transform, visualize, and model their datasets. "Key Accomplishments:" --Data Cleaning: Implemented functionalities for handling missing values, removing duplicates, and converting data types. --Feature Engineering: Enabled users to create new features, select relevant ones, and generate polynomial features to enrich their datasets. --Data Transformation: Applied log transformation, polynomial features, and binning techniques to prepare data for modeling. --Model Building: Facilitated training and evaluation of machine learning models, including hyperparameter tuning and cross-validation. --Interactive Visualizations: Integrated Plotly for generating a wide range of interactive plots such as histograms, scatter plots, bar charts, heatmaps, and pie charts. --Download Modified Data: Provided options for users to download cleaned, transformed, and feature-engineered datasets. --Technologies Used: Python, Streamlit, Plotly, pandas, numpy, scikit-learn GitHub: https://github.com/Divyansh0108/DataAnalysisToolkit This project showcases advanced capabilities in data preprocessing, feature engineering, and visualization, offering an all-in-one solution for comprehensive data analysis and making the data science workflow more accessible and efficient.

Comprehensive Data Analysis To

This project involved developing a comprehensive text summarization system with a focus on the full MLOps lifecycle, from CI/CD integration to cloud deployment. Key Accomplishments: -CI/CD Pipeline: Automated deployment with GitHub Actions for efficient and error-free builds. -Dockerized App: Ensured consistent environment replication and deployment using Docker. -AWS Deployment: --EC2: Hosted the application on scalable and secure virtual machines. --ECR: Stored Docker images for easy access and deployment. -Model Evaluation: Assessed summarization quality using ROUGE and SacreBLEU metrics. -Technologies Used: --NLP Libraries: Transformers, NLTK --Tools: FastAPI, Docker, GitHub Actions, AWS EC2/ECR --Focus of the Project: The main goal was to develop a production-ready pipeline with a seamless development, testing, and deployment process, rather than focusing solely on the model. -GitHub Repository: https://github.com/Divyansh0108/E2E-text-summarization The project highlights expertise in full MLOps life cycle, including CI/CD, containerization, and cloud deployment, demonstrating the ability to manage and deploy robust applications.

End-To-End Text Summarization 

Developed a machine learning-based heart disease prediction model and deployed it as a web application. This project aimed to leverage data science techniques to assist in early detection and risk assessment of heart disease. Key Accomplishments: - Implemented and compared multiple machine learning algorithms including Logistic Regression, Random Forest, Gradient Boosting, and SVM - Achieved 0.852459 test accuracy with the best performing model (Logistic regression) - Developed an interactive web application using Streamlit for real-time predictions - Utilized data preprocessing techniques and implemented a robust model training pipeline - Applied data visualization techniques to analyze model performance and dataset characteristics Technologies Used: Python, scikit-learn, pandas, numpy, Streamlit, matplotlib, seaborn GitHub: https://github.com/Divyansh0108/heart-disease-pred This project demonstrates proficiency in machine learning, data preprocessing, model evaluation, and web application development, showcasing the potential of AI in healthcare applications.

Heart Disease Prediction Model

Developed an AI-driven tool to predict and assess mental health status, specifically targeting depression among students. This project leverages machine learning to identify depression likelihood based on diverse input features, offering personalized insights and actionable recommendations. It provides an accessible, transparent, and proactive solution for mental health management, benefiting students, counselors, and educational institutions. Key Accomplishments: Early Detection: Built a neural network model using TensorFlow, trained on a dataset of 27,000 student records, to accurately predict depression risks before symptoms escalate. Personalized Insights: Designed a system to analyze factors like sleep duration, academic pressure, and dietary habits, delivering tailored recommendations for individual users. Model Interpretability: Integrated LIME to provide transparent feature importance analysis, enabling users to understand prediction reasoning. User-Friendly Interface: Created an intuitive Streamlit-based web application for seamless data input and result visualization, offering both quick and detailed analysis modes. Proactive Support: Developed a system to provide actionable next steps, including self-care strategies and professional consultation prompts, fostering preventive mental health care. Data Preprocessing: Implemented robust preprocessing steps, including handling missing values, feature encoding, and data filtering, to ensure high-quality input data. Technologies Used: Python, TensorFlow, Streamlit, Google Generative AI, Matplotlib, LIME, scikit-learn, pandas, NumPy

Mental Health Predictor and Ad

Developed a Note Taking App using Django and associated technologies to enhance user productivity and note management. This project aimed to provide a comprehensive solution for users to manage their notes effectively with an intuitive interface. Key Accomplishments: Designed and implemented a user-friendly interface for creating, viewing, editing, and deleting notes. Integrated core functionalities including user authentication, note management, and a rich text editor. Added features such as search and filtering, file uploads, and collaboration tools to enhance usability. Ensured smooth app performance and security through the use of Django, Pillow, Whitenoise, and SQL for backend functionality. Technologies Used: Django, Pillow, Whitenoise, SQL, Bootstrap, JavaScript GitHub: https://github.com/Divyansh0108/Note-Taking-Python-Web-Tech This project demonstrates proficiency in web development with Django, full-stack development, and the implementation of user-centric features, showcasing the ability to build and deploy interactive web applications.