Projects

Research and engineering work, mostly in computer vision and deep learning.

Prism.ai

A six-stage AI research pipeline that takes a raw dataset from analysis through literature retrieval, experiment planning, and code generation to a written research report.

• PyTorch
• FastAPI
• Next.js
• OpenAlex
• GPT-5.6

PrepArt

Final year project: a real-time multimodal mock-interview system that fuses facial and eye cues, vocal prosody, and transcript analysis into structured interview feedback.

• Python
• MediaPipe
• Librosa
• MoviePy
• PyTorch
• FastAPI

Brain Tumor MRI Classification

Four-class brain tumor classification from MRI using MobileNetV2 transfer learning, with Grad-CAM used to check that the model attends to tumor regions rather than scan artifacts.

• PyTorch
• MobileNetV2
• Grad-CAM
• scikit-learn
• NumPy

Skin Lesion Classification

Seven-class dermatoscopic lesion classification on HAM10000 (~10k images) with ResNet-18 transfer learning and class-weighted loss to counter severe class imbalance.

• PyTorch
• ResNet-18
• HAM10000
• torchvision
• scikit-learn