Live at sowaibaarshad.com
I am an AI/ML Engineer and Software Developer specializing in Large Language Models, Retrieval Augmented Generation systems, and computer vision. Silver Medalist in Software Engineering at UET Taxila (CGPA 3.94/4.00). This repository is my portfolio website, built with Next.js 14 and React.
The portfolio does not just describe my work, it runs it:
- ThoraxNet, live: upload a chest X-ray and my deployed BioMedCLIP (ViT-B/16) model returns probabilities for 14 pathologies with MC-Dropout uncertainty and an auto-generated radiology report, served from a Hugging Face Space
- DeepGuard, live: upload a face photo and my fine-tuned EfficientNet-B4 decides real or fake, with a GradCAM heatmap showing where the model looked, plus one-click samples from my own Deepfake dataset
- Ask-My-Portfolio RAG: a TF-IDF retrieval engine with cosine ranking over my profile, written from scratch and running entirely in the visitor's browser
- 3D lab carousel: drag-to-rotate card ring with pointer physics, snap animation, and a modal system that preserves component state
- Live GitHub stats in the boot sequence, light and dark themes, scroll reveal, keyboard navigation, and reduced-motion support
Next.js 14 (App Router), React 18, plain CSS with custom properties for theming. No UI libraries. The retrieval engine and carousel math are hand-written modules in lib/ and components/Lab.jsx.
app/ layout, page, global styles
components/ StatusBar, Hero, About, Skills, Experience, Projects, Lab, Education
components/lab/ ThoraxCard, DeepGuardCard, AgentCard, RagCard
lib/ rag.js (TF-IDF engine), github.js (cached profile fetch)
npm install
npm run dev- ThoraxNet · medical AI framework for chest X-ray pathology detection
- DeepGuard-XAI · deepfake detection with explainability
- ClinicaQuery-AI · RAG over clinical papers with source attribution
- Devops-swarm · autonomous self-correcting coding agent (LangGraph)
- AgroVision-Net · plant disease classification across 38 classes