Omer – Python, Scikit-learn, Keras

Covid-19 diagnosis from X-rays with Transformers

• Obtained >99% Covid-19 detection from x-rays by using transformers and transformer-cnn hybrids.
• Ran training experiments through Colab with the latest Covid-19 data.
• Compared the results of transformer, cnn and transformer+cnn architectures.
• Found that the best architectures for Covid-19 detection are Swin Transformer and CoatNet.

UI element detection and screen classification

• Created object detection models for 30+ UI elements with >90% mAP
• Implemented an active learning workflow with Colab and Labelbox
• Integrated modeling runs with Weights & Biases for model, data and experiment tracking & versioning
• Implemented a workflow for MAL (model assisted labeling) in Labelbox
• Trained and deployed the UI object detection model with AWS Sagemaker
• Investigated UI understanding by deploying a mid sized vision LLM model

Skin cancer is one of the most dangerous forms of cancer. Skin cancer spreads over other body parts gradually, so it is more durable in the initial stages, which is why it is best detected at the early stages. Omer was part of a team that developed a highly accurate skin cancer detection deep learning model. The model lets users classify skin lesions as benign or malignant with very high accuracy.

Main accomplishments:

• Developed deep learning CNN networks to identify skin cancer from photos with more than 95% sensitivity and specificity;
• Was in charge of data cleaning, model development, and model tuning.

Advanced machine learning prediction model for COVID-19 prognosis combines several variables or features to estimate the risk of health outcomes for patients infected with COVID-19 and to assist medical staff in triaging patients with limited healthcare resources.

• Developed a COVID-19 prognosis model with an average of 80% accuracy;
• The model lets users predict health outcomes for patients days in advance after they were hospitalized;
• Was in charge of data analysis, data cleaning, feature engineering, and model selection.

The platform for autonomous machines and other embedded applications. It provides customers with the performance and power efficiency they need to build software-defined autonomous machines and makes it possible to bring new capabilities to millions of tiny, low-power AI systems.

• Developed an embedded machine vision system to detect impurities in cotton with Nvidia Jetson;
• Set up the system on factory premises with a siren and a flashing light;
• Researched how to detect white contaminants;
• Was in charge of data collection, model development, deployment, and testing;
• The final tests of the project were 100% successful.

As the most popular technology in the 21st century, machine learning can help realize many gains in healthcare, including better diagnosis, care, and treatment. Therefore, applying AI-assisted care to clinical medicine is vital, especially in the context of COVID-19 pandemic prevention and control work. Omer, together with a team, has developed a COVID-19 diagnosis tool from blood tests with more than 90% sensitivity and specificity.

• Created a machine learning model for COVID-19 diagnosis with more than 90% accuracy and deployed it on the web;
• Built the backend (API) with Flask;
• Retrained the model with more incoming data;
• Wrote a journal paper;
• Was in charge of data analysis, data cleaning, feature engineering, model selection, and model deployment.

Chest x-rays are a fast and inexpensive test that may potentially diagnose COVID-19, the disease caused by the novel coronavirus. However, chest imaging is not a first-line test for COVID-19 due to low diagnostic accuracy and confounding with other viral pneumonia. Recent research using deep learning may help overcome this issue as convolutional neural networks (CNNs) have demonstrated high accuracy of COVID-19 diagnosis at an early stage. Omer was one of the Data Scientists that worked on developing a deep-learning model for COVID-19 diagnosis from X-ray data. The model can diagnose COVID-19 from X-ray pictures, therefore either replacing or confirming PCR tests.

• Created a deep learning model for COVID-19 diagnosis from X-ray data with 92% average accuracy;
• Researched several models from literature and selected the best one;
• Was in charge of model development and selection.