Projects
Face Generation: What's Under the Mask
Python, Google Colab, TensorFlow, Keras, Matplotlib, Seaborn, GAN, Deep Learning
- Analyzed 200k CelebFace images and developed new masked image dataset by extracting facial landmark data.
- Implemented GAN model using Deep Learning techniques for accurate mask object detection and image generation, achieving 90% accuracy in distinguishing real/fake images. Libraries used include Matplotlib, Seaborn, TensorFlow, and Keras.
Walmart Sales Analytics Dashboard
Python, MS Power BI, ETL, Data Visualization
- Built an interactive analytics dashboard of State-wise sales performance of Walmart in the USA using multiple drill-down filter options. This involved using ETL (Extract, Transform, Load) and experimentation findings for a better understanding of sales and profit by orders' year, category, and more, to deliver business insights into customer experience, ultimately reducing manual prediction time by 90%.
Prediction of Bike share users per hour
Python, Scikit-Learn, Predictive Modeling, Regression Analysis, Data Visualization, Google Colab
- Engineered a predictive model to forecast hourly Capital Bikeshare demand, processing over 17,000 data points to achieve a 76% validation accuracy.
- Benchmarked regression algorithms, identifying that K-Nearest Neighbors (KNN) outperformed linear models in capturing non-linear correlations between weather severity and rental volume.
- Optimized model performance by conducting rigorous exploratory data analysis (EDA), feature scaling, and outlier detection on meteorological and temporal variables.
Power to the People: Lumos
Android, Mobile Development, Real-time Location Services
- Designed and developed a user-friendly Android application that connects local communities to share essential resources such as generators, food, and shelter during power outages, promoting resilience and mutual aid.
- Implemented real-time location-based matching and resource status updates, ensuring users could quickly find and offer help, reducing response time in emergencies.