Project Gallery
A curated selection of technical research architectures and automated software solutions.
Machine Learning & Deep Learning
Liver Cancer Risk Prediction (ML Pipeline)
Reproducible ML workflow for liver cancer risk prediction from clinical and lab data. Includes preprocessing and clinical feature engineering (age groups, BMI, comorbidity, liver-function index), categorical encoding, and training of multiple classifiers. Evaluated with accuracy, ROC‑AUC, precision, recall, and F1; outputs include tables and plots (confusion matrices, coefficients, feature importance). Configurable CSV inputs for easy adaptation and deployment.
Iris Dataset Exploratory Data Analysis
Executed a thorough exploratory analysis of the Iris dataset: data cleaning and validation, summary statistics, boxplots and histograms for feature distributions, pairwise scatterplots and PCA for dimensionality insights, and correlation analysis to detect multicollinearity. Findings highlighted key feature separations among species and informed feature selection and modeling strategies.
Chronic-Kidney-Disease-Risk-Factor-Prediction-Using-Machine-Learning
Built a machine-learning pipeline to predict chronic kidney disease risk from clinical and lab data. Tasks included data cleaning, feature engineering, model selection (e.g., logistic regression, random forest, XGBoost), cross-validation, and evaluation with ROC/AUC and precision-recall metrics. The project identified important predictors and produced an interpretable risk score to support early detection.
Alexa-Simplified virtual assistant
I built a simplified Alexa-based virtual assistant focused on core tasks like voice queries, scheduling, reminders, and basic automations. The design emphasizes lightweight, reliable performance with streamlined intent handling and minimal integrations for fast response times and easy deployment.
Life_span_age_transformation
Jupyter notebook demonstrating lifespan–age transformation methods to prepare longitudinal biomedical datasets for analysis. Includes data cleaning, age normalization and rescaling, feature extraction, visualization of age-related trends, and examples of downstream modeling workflows for predicting health outcomes.
Real-time object detection
Real-time object detection for live video using a compact CNN, NMS, and lightweight tracking to output fast, accurate bounding boxes and labels. Optimized for edge deployment via pruning, quantization, and a streamlined preprocessing pipeline.
Age-Gender-Detection
I built a lightweight age-and-gender detector using deep learning and transfer learning. The system detects faces, preprocesses and augments images, and trains on balanced data. After evaluating accuracy and error metrics, I deployed the model as a fast, real-time service optimized for low-resource devices.
Invisible Cloak
Implemented a real-time invisible-cloak system that detects a colored cloak, captures a background reference, and substitutes cloak pixels with the background to produce an invisibility effect. The pipeline employs color-space segmentation, frame-wise masking, and blending, and is optimized for low-latency performance on consumer hardware.
Flappy-bird
I built a Flappy Bird clone with a simple game loop, tap/space controls, gravity-based flap physics, moving pipe obstacles with randomized gaps, collision-based game over, and score tracking. I optimized performance with object pooling, lightweight assets, and separate rendering/logic for a smooth, responsive experience on web and mobile.