At IGDTUW, I developed a Flutter mobile application integrated with TensorFlow Lite for real-time handwritten digit recognition, achieving 97.6% accuracy. I also engineered an interactive drawing canvas and optimized preprocessing, reducing input processing time by 25%.
In a Python & ML internship at IGDTUW, I built time-series forecasting models with Python and Scikit-learn using financial data. Feature engineering and statistical analysis helped identify trends in stock market datasets.
For my Handwritten Digit Recognition project, I created a browser-based system with an HTML5 canvas and a real-time JavaScript classification pipeline. I optimized the model to reduce inference latency by 20%.
I also developed a customer segmentation model using historical transaction data and cosine similarity to group users with similar behavior. I visualized segmentation patterns with Matplotlib, and co-authored a paper on handwritten digit recognition published in Springer SMARTCOM 2025.
