Featured Projects

Tech Stack

Description

SMARTDERMSCAN is an AI-powered skin cancer detection system that uses deep learning, computer vision, and explainable AI to help with early screening and clinical decision support. The system is trained using two major public datasets: HAM10000 and ISIC 2019, which contain labeled dermoscopic images of different skin lesions.

It supports multi-modal detection, meaning it can analyze single images, recorded videos by extracting multiple frames, and live camera feeds in real-time. The system performs both segmentation and classification, where segmentation includes hair removal, lesion masking, and lesion highlighting to improve accuracy, while classification currently supports seven types of skin lesions with a clear plan for future expansion.

It also includes an offline AI doctor assistant powered by local models such as BioGPT and Chat Doctor, which helps users understand diseases and diagnostic results without sending any data to external servers, ensuring full privacy and security. This makes SMARTDERMSCAN suitable for real-world clinical use, education, and remote screening where internet access and data privacy are serious concerns.

Key Features

• Multi-modal detection (images, videos, live camera)
• Advanced segmentation with hair removal
• Classification of 7 types of skin lesions
• Offline AI doctor assistant for privacy
• Explainable AI for clinical decision support

Tech Stack

Description

The multimodal intelligent robot arm is an advanced AI-driven system designed to perform tasks through multiple interaction modes. It supports custom model uploads, allowing users to integrate their own machine learning models for specialized operations.

The arm can be controlled through a chatbot interface, voice commands, or manual input, offering flexibility and ease of use. It includes forecasting capabilities using ARIMA and LSTM models for predictive tasks such as demand or motion prediction. Additionally, it features object detection powered by computer vision, enabling accurate identification and manipulation of real-world objects.

Key Features

• Chatbot, voice, and manual control
• Custom ML model integration
• Forecasting with ARIMA and LSTM
• Computer vision object detection
• Real-time robotic control

Tech Stack

What's Unique

Shadow OS is unique because it is built from zero without using Linux or Windows code. It uses a hybrid kernel which mixes monolithic and microkernel ideas. It includes secure folder access with sandboxed file system isolation and an AI security agent that monitors system calls and blocks suspicious behavior. It is 64-bit from the start and focuses on real hardware drivers like keyboard, mouse, and framebuffer graphics.

Technologies Used

• GRUB2 bootloader for booting the system
• Multiboot2 specification for kernel loading
• x86-64 CPU architecture
• Paging and virtual memory for RAM management
• Programmable Interrupt Controller and timer for multitasking
• Framebuffer graphics for GUI
• PS/2 keyboard and mouse drivers
• ELF64 binary loader for programs
• Make and Clang toolchain for building the OS

Programming Languages

C: Kernel, OS core, memory management, and drivers

x86-64 Assembly: Boot code, CPU setup, and context switching

Shell scripts and Makefile: Build automation