Saied Salem
Machine Learning Engineer
Biography
I’m Saied Salem, a recent graduate from Faculty of Engineering at Cairo University, Egypt. and Ex- R&D engineer at Astute Imaging.
My area of interests includes computer vision, image processing, NLP, Deeplearning, Scientific machine learning , Digital signal processing and software engineering. with a GPA of 3.7 out of 4.0 and 6.5 IELTS score
Interests
- Machine learning
- Computer vision
- Natural language processing
Education
B.Sc Systems and Biomeical Engineering, 2023
Cairo University
Experience
Research and Development Engineer
- Contributing to developing an End to end PACS-integrated system with automatic segmentation and classification workflow of breast cancer ultrasound images using state-of-the-art deep learning techniques, enhancing the efficiency of the diagnosis
- Developed a DICOM medical viewer with various measurements and image processing tools for images and videos acquisitions, aiding physicians in their assessments
- Implemented a configurable breast ultrasound segmentation package for training deeplearning models, achieving the highest dice score in the literature with new hyperparameter optimizations
- Utilized Azure Virtual Machines (VM) to host and run the containerized system.
Data scientist intern
- Working on developing human input inspection system, creating data acquisition pipline and comparing it with deeplearning field detecting system api
- building machine learning model for companies credit limit estimation
Projects
*
Developed an end-to-end PACS-integrated system for the automatic segmentation and classification of breast cancer ultrasound images using state-of-the-art deep learning techniques
Configurable classification package for 3D lung Nodule CT that incorporates MLOps and Automated-parallel hyper parameter tuning
Fully customized 3d brain tumors segmentation engine that incorporates traditional approaches such as Z-net and DeepLabv3+ and probabilistic approaches like attention Unet with VAE
A 3D medical viewer built with vtk-js that supports volume rendering with multiple presets and marching cubes
Neural machine translation model utilizing LSTM with attention mechanism to convert various date formats into a standardized format
Built a speech dataset by employing DSP concepts, including histogram and FFT, and then utilized a GRU-based model for trigger word detection
The application leverages Z-transform concept to construct filters with zeros and poles, capturing the phase and magnitude response then filters real-time input signals and dynamically plots the results in real-time
Implementing Temprature displayer using ADC drive with LM35 tempreture sensor and LCD using STm32 microcontroller (ARM)
A multi-threaded application for curve fitting and interpolation to demonstrate the best way of fitting data by using one or more chunks and a graphical map for the error
Applying DSP concepts for spectrum analysis and instrument manipulations within songs with dynamic visualization
Design RTOS-based implementation that employs real-time design patterns for a vehicle direction and hazard indicator control system, effectively managing the vehicle indicator LEDs
Developing an illustrator for the signal recovery that shows Nyquist rate. read csv signal and see the sampled points highlighted on top of the signal. Change the sampling rate via a slider that range from 0 Hz to 3f max Reconstruct/recover the signal from the sampled points. Application has a Composer to generate basic signals to test and validate on the app One graph to display the sinusoidal to be generated One graph to display the sum of the generated sinusoidals. A combobox to select one of the contributing sinusoidals and remove it via a delete button. After making a synthetic signal then moving it to the main illustrator graph to start the sampling/recovery process