Coding Research Projects
Self-Supervised Learning
Explored contrastive methods (SimCLR, NNCLR, DINO, SimSiam, BYOL) to build rich representation spaces and perform nearest-neighbour retrieval.
Conditional Invertible Neural Networks
Built CI-NNs with Glow and spline coupling blocks: first using tabular scalars to infer target properties, then using learned embeddings with a mixture-of-experts.
Personal Projects
InvoiceSense
Python end-to-end document understanding pipeline on the Ghega dataset: OCR (Tesseract), rule-based NLP field extraction, and precision/recall evaluation.
Sign Language
In-progress ML web app trained on CIFAR: upload an image and get back class probability predictions from a deep network.
Portfolio Website
This very site—hand-crafted with PHP, HTML & CSS. Includes templating, optional auth, responsive design, and project showcases.