Crafting Intelligent AI Systems
with Deep Learning & Cloud.

A medical AI that actually verifies its answers — not guesses them. 10 specialist AI agents backed by 125K+ trusted medical documents. Fine-tuned Google MedGemma for cardiology. HIPAA & GDPR compliant.

• 10 AI agents collaborate like specialist doctors — each handles a different type of medical question.
• Self-correcting RAG pipeline verifies every answer against 125K+ medical documents before responding.
• 55K+ lines of Python across 220+ files — fine-tuned MedGemma with HIPAA-grade security.

A comprehensive fintech ecosystem with real-time transactions, AI fraud detection, loan eligibility prediction, and an immersive 3D Three.js login — all on a double-entry ledger for financial integrity.

• ML models for fraud detection (TF-IDF + Logistic Regression) and loan risk prediction.
• Double-entry ledger system guarantees financial data integrity — no balance drift.
• 3D immersive auth page with Three.js — distinctive UX that sets it apart from template banking apps.

Handles 1GB PDF files without freezing the UI. Celery + Redis task queues offload heavy OCR and compression. Real-time WebSocket progress. AES-256 encryption. Dual interface: Flask web UI + FastAPI REST.

• Async task queues (Celery + Redis) handle large-file OCR and compression without blocking.
• 25+ PDF operations — merge, split, compress, convert, rotate, password-protect, and OCR.
• Real-time progress via WebSockets; AES-256 encryption for sensitive document workflows.

Full-stack appointment and patient management platform with conflict-free scheduling, multi-role access control, and a secure REST API — built with Node.js and PostgreSQL.

• Optimized PostgreSQL scheduling logic using constraint checks to prevent doctor double-booking.
• Role-Based Access Control (RBAC) with separate permission sets for doctors, patients, and admins.
• JWT auth with refresh token rotation and Helmet.js + rate-limiting middleware for API security.

A production-quality collection of 34 Jupyter notebooks across machine learning, deep learning, computer vision, and image processing — built during academic study with real datasets.

• Heart Disease Prediction: XGBoost + LightGBM + SHAP explainability + Optuna hyperparameter tuning.
• Transfer Learning with VGG16, ResNet50, InceptionV3, and MobileNet on custom datasets.
• Computer Vision: edge detection, LBP texture analysis, image similarity search, and segmentation.