• "Robot LUNA", a waiter robot, uses a dual camera setup (wide-angle camera and a Kinect-2 depth camera) for 3D point cloud-based navigation in restaurants. Its enhanced stability circuits ensure safe food delivery without spillage. Internally, it uses three Raspberry Pis for parallel processing and an Atmega2560-based custom PCB to get sensor/encoder readings and to control motor drivers. It communicates with a restaurant’s Computer server with ROS 1 Noetic via local wifi for path planning and precise navigation to order locations. This makes LUNA a reliable and efficient addition to the restaurant staff.

• Tools & technologies used: ROS 1 Noetic, TensorFlow, OpenCV, Open3D, C++ with Atmega2560 custom PCB for Motor and stability Control, Python with Raspberry Pi 4b - (3 SBCs), Kinect v2 depth camera

-BioSense-AI: ECG-PPG hardware and a Mobile App- https://github.com/BioSense-AI - 2023 - present

• Pocket-size electronic device with a Mobile app to display and analyze Body signals with a Multi-Model 2
• Project BioSense-AI is a both software - hardware project focusing on predicting diseases using ECG-PPG, temperature, and text prompts. Signals are amplified and filtered via our custom analog circuits, then fed to Orange Pi Zero 2W SBC from an ADC to digital processing. The system uses a deep learning custom multi-model architecture with interpretability for accuracy and efficiency. Currently, we are developing a mobile app for user access to the AI model by scanning ECG reports and other prompts for hospitals without our electronic device.
• Tools & technologies used: PyTorch, GradCAM, lime, Analog Filter Design, Setting up Orange Pi zero 2W SBC for custom tasks, I2C protocol, ADC, Flutter

-HyperTalk-Computer-Vision: Deep learning-based Sign Language translating Mobile App- https://github.com/HyperTalk-Computer-Vision - 2022 - present

• Real-time bidirectional sign language translating AI tool for deaf and speech-impaired individuals 2
• HyperTalk mobile app and the website both are computer vision-based solutions to facilitate communication in sign language for individuals with hearing and talking issues. Which is capable of real-time translation in both directions:

1] Sign language camera feed to Voice.

2] Voice feed to sign language animations.

• Ongoing focus is on enhancing the first feature with a new continuous Word-level Sign Language Recognition model that is capable of more accurate and faster translations with different sign language options for different regions in the world - Based on the following Data sets:

Phoenix 2014 Dataset (German Sign Language Videos)

OpenASL Dataset (American Sign Language Videos)

CSL Dataset (Chinese Sign Language Videos)

BOBSL Dataset (British Sign Language Videos)

• Tools & technologies used: PyTorch, CUDA, ONNX, Flutter, OpenCV, Django back-end development

-Smarty-Plug: IoT smart extension cord- https://github.com/Smarty-Plug - December 2022

• IoT-enabled extension cord with smart controls, offering voice commands and scheduling 2
• Smarty Plug is an IoT-integrated smart extension cord powered by ESP8266 and Atmega328 which is designed with the following features:

Voice Controlling - supports Google Assistant and Amazon Alexa

Controlling by Mobile Phone - using Google Home

Scheduling - by connecting to Google Calendar

USB Adaptive Charging

• Tools & technologies used: Altium PCB designing, Atmega328p and ESP8266 coding, Solid Works, Platform IO

-Cosmo-Robot; https://github.com/Cosmo-Robot May 2023

• Designed to showcase various capabilities with a robot hand, Arduino Mega 2560, and Robotic sensors 2
• "Cosmo" robot is designed to showcase various capabilities, including precise line following, obstacle avoidance during line following, navigating ramps at 20 - 30 degrees, interacting with objects using a mechanical arm, sound sensitivity, color detection, and maze-solving capabilities. Additionally, it was programmed to avoid other robots in the arena such as the guard robot, on its way to the final destination.
• Tools & technologies used: Platform IO, Arduino Mega 2560 coding, Robotic sensors