Explore machine learning models and cloud-deployed apps I've built, covering NLP, computer vision, and cloud deployment workflows.
A Transformer model built from scratch in PyTorch and trained on WMT16. It includes multi-head attention, positional encoding, subword tokenization, and beam/greedy decoding. Hosted on a GCP virtual machine and served via Flask, this project features cross-attention visualizations and token alignment graphics.
A deep convolutional neural network trained on the NIH Chest X-ray14 dataset to detect common thoracic diseases from radiographic images. The model is deployed via an EC3 GPU instance and accepts image input directly from the user for inference. Frontend integrates with Flask for upload, display, and prediction.
A lightweight Random Forest classifier trained on the Iris dataset and deployed using AWS SageMaker. Demonstrates end-to-end ML deployment, including input handling, REST API usage, and live prediction through Flask integration.
This website was built from scratch using Flask, Jinja2, and custom HTML/CSS. It integrates live demos, attention visualizations, and project write-ups. Hosted on Render, it serves as both a portfolio and a real-time ML testing interface.
Future additions include a BERT-based sentiment classifier, a YOLOv8 object detection demo, and a full MLOps pipeline with CI/CD.