
Abid ilyas
dev
Skills

Bekijk mijn diensten


Werkervaring
Controlling system
OUSSAMA • Freelance
Oct 2024 - Present • 1 yr 7 mos
I am a 2024 Software Engineering graduate and Independent Consultant with a track record of building high-impact digital systems. Rather than following a traditional path, I have spent my early career solving complex problems for clients by bridging the gap between advanced logic and practical usability. My expertise is built on a versatile stack of Python, Java, JavaScript, and Android development. I specialize in creating 'smart' systems—ranging from AI-driven medical diagnostic tools that analyze X-rays for bone fractures to integrated control platforms that sync web and mobile apps to optimize business operations. I don’t just write code; I design solutions that 'manage time on the good side,' focusing on automation and efficiency. Whether it’s developing a backend in Python or a native Android experience, I bring fresh energy, modern technical standards, and a commitment to delivering clean, maintainable software that solves real-world challenges. The Technical Challenge Medical imaging presents unique challenges compared to standard object detection. X-rays often contain "noise," varying contrast levels, and subtle fractures that are easily overlooked. My task was to build a pipeline that could: Pre-process raw DICOM/JPG images to normalize lighting and contrast. Isolate anatomical regions (e.g., distinguishing a wrist from an ankle). Identify discontinuities in bone structure that indicate a break. System Architecture & Development I chose Python as the core language due to its robust ecosystem for scientific computing. The system was built using a multi-stage pipeline: Image Pre-processing: I utilized OpenCV to implement Otsu’s Binarization and Canny Edge Detection. This helped in stripping away "non-bone" visual data and highlighting the edges of the skeletal structure. Feature Extraction: Leveraging TensorFlow and Keras, I implemented a Convolutional Neural Network (CNN). I focused on a ResNet-50 architecture, which is highly effective at recognize