Selected work spanning MLOps, machine learning, and full-stack applications.
Deployed a production-ready Heart Disease Prediction API on Google Kubernetes Engine (GKE) using Docker and FastAPI. Configured HorizontalPodAutoscaler (HPA) for load management and implemented comprehensive observability via Google Cloud Logging. Integrated SHAP for model explainability and performed fairness analysis (Fairlearn) to detect gender bias.
Participated in Kaggle's Crime-Cast competition as part of an IIT Madras project, building and optimizing machine learning models using Python, scikit-learn, and XGBoost to predict crime categories. Applied feature engineering, data preprocessing, and visualized insights using Seaborn and Matplotlib.
Developed a comprehensive web application featuring Role-Based Access Control (RBAC) for librarians and users, built with Flask, Vue.js, and SQLite. Implemented Redis caching for performance optimization and utilized Celery for asynchronous background jobs, such as daily reminders and monthly activity reports.
Built an image captioning system leveraging a CNN-LSTM architecture with VGG-16 for image encoding. Improved performance with Batch Normalization and Dropout techniques and evaluated results using BLEU scores. Developed and trained the model using TensorFlow/Keras.