I am a Software Developer and Data Scientist with a strong foundation in building and deploying end-to-end intelligent systems. My expertise spans both full-stack web development and the complete MLOps lifecycle, from data versioning and experiment tracking to containerization and CI/CD.

I am passionate about creating robust, data-driven applications that solve real-world problems, with core interests in Machine Learning, Computer Vision, and Natural Language Processing.

• Languages & Frameworks: Python, JavaScript (Node.js, React), TensorFlow, Keras, Scikit-learn, Pandas, NumPy, Langchain, C++.
• MLOps & Deployment: DVC, MLflow, Docker, GitHub Actions (CI/CD), Flask, Express.js
• Databases: SQL
• Deep Learning:
  • Computer Vision (CV): Convolutional Neural Networks (CNNs), Image Classification, Transfer Learning
  • Natural Language Processing (NLP): Recurrent Neural Networks (RNNs), LSTMs, Retrieval-Augmented Generation (RAG), Sentiment Analysis

A full-stack web application designed to streamline the process of submitting, tracking, and resolving grievances.

• Architecture: Built with a React.js frontend and a Node.js (Express) backend.
• Database: Uses a SQL database hosted on AWS for data persistence.
• Core Features: Secure user authentication (JWT), role-based access control (User, Office Bearer, Approving Authority, Admin), grievance submission with file uploads, and a multi-level grievance escalation system.
• Notifications: Integrates email notifications for status updates and OTP verification.

A complete MLOps pipeline for classifying chest cancer from X-ray images. This project demonstrates a production-level workflow.

• Data & Model Versioning: Utilizes DVC to manage datasets and large model files efficiently.
• Experiment Tracking: Integrates MLflow for logging experiments, tracking metrics, and managing model versions.
• Deployment: The application is fully containerized using Docker and includes a CI/CD pipeline with GitHub Actions for automated building and testing.
• Application: Served via a Flask backend that provides a simple web interface for predictions.

A Retrieval-Augmented Generation (RAG) chatbot designed to answer user questions about a specific YouTube video.

• Architecture: Implements a RAG pipeline that leverages a vector database to retrieve relevant information from the video's transcript.
• Components: Built with a server-side backend to handle the NLP logic and a browser extension frontend for seamless user interaction directly on the YouTube page.

I believe in continuous learning and maintain public repositories to document my knowledge. ([https://github.com/kumathepanda/deep-learning-journey])

• Deep Learning Journey: A comprehensive collection of notebooks implementing and explaining core deep learning concepts, including:
  • Artificial Neural Networks: Backpropagation, Gradient Descent, Hyperparameter Tuning.
  • Convolutional Neural Networks: CNN architectures (LeNet), Dog vs. Cat image classification, transfer learning.
  • Recurrent Neural Networks: LSTMs for next-word prediction and sentiment analysis.
  • Keras Functional API: Building complex models with multiple inputs and outputs, such as an age and gender classifier.

([https://github.com/kumathepanda/machine-learning_journey-part1]) ([https://github.com/kumathepanda/machine-learning_journey-part2])

• Machine Learning Journey (Parts 1 & 2): A detailed log of projects and notebooks covering fundamental machine learning algorithms from scratch and with Scikit-learn.
  • Models: Linear/Logistic Regression, K-Means, DBSCAN, Agglomerative Clustering, Decision Trees, Random Forests, and Gradient Boosting.
  • Techniques: In-depth EDA, advanced data preprocessing, feature engineering, and dimensionality reduction with PCA.

• LinkedIn: [LinkedIn Profile]
• Email: [ramdeoparth@gmail.com]