Real projects built with Python, JavaScript, and deep learning — from geological visualization to handwritten math recognition.
A geological cross-section renderer with stratigraphic correlation, borehole data visualization, and interactive layer management for fieldwork professionals.
Built using a custom 2D canvas rendering engine to draw dynamic Bézier curves that connect stratigraphy across multiple boreholes. The data is managed via a React frontend with a modular architecture.
React, JavaScript, HTML5 Canvas, TailwindCSS
OCR classifier that recognizes handwritten mathematical expressions using a custom CNN pipeline trained on symbol datasets with real-time inference.
Designed a Convolutional Neural Network (CNN) architecture optimized for spatial hierarchy recognition. Implemented data augmentation techniques to handle varying handwriting styles.
Python, TensorFlow, Keras, OpenCV, NumPy
Deep learning based digital watermarking system using convolutional neural networks for embedding and extracting imperceptible watermarks in images.
Utilized an encoder-decoder network architecture to invisibly embed watermark tensors into image spatial domains, maintaining robustness against compression and filtering.
Python, PyTorch, Matplotlib
A cafe management and order tracking system built with JavaScript for streamlining daily operations, inventory, and customer workflows.
This portfolio website — a cinematic 3D moon experience with Three.js, scroll-driven animations, and interactive hover effects.
A Python calculator demonstration project exploring basic computation logic and interface design.