I build practical AI systems, including document intelligence pipelines, and Im looking to connect with other AI practitioners and teams.
Portfolio highlights systems and pipelines for document intelligence, OCR, data extraction, and retrieval-augmented generation applied to mortgage files.
Outamation · Feb 2026 · ✅ Verified by Extern
An Outamation externship on AI-powered document intelligence, focused on turning unstructured mortgage files into searchable data through OCR, extraction, and RAG pipelines.
Built an end-to-end fraud detection pipeline on 1.3M+ transactions, training on Random Forest, XGBoost, and Logistic Regression to achieve a 99.53% accuracy and 0.97 ROC-AUC. Mitigated class imbalance using SMOTE and identified 15 high-impact features via SHAP analysis, boosting precision from 45.23% to 57.89% while improving model explainability. Collaborated with a team of 4 to optimize classification threshold and conducted 10-fold cross-validation to boost recall to 66.76% on a heavily imbalanced dataset.
I build practical AI systems, including document intelligence pipelines, and Im looking to connect with other AI practitioners and teams.
I am Mohamed Awad, a senior Computer Science student at CUNY Queens College pursuing a bachelors degree. I focus on artificial intelligence and have completed an externship building AI-powered document intelligence for mortgage files. I am looking to grow my network and connect with peers and mentors in AI.
Externships
Outamation Advanced AI-Powered Document Insights and Data Extraction Externship
Completed · Feb 2026
Experience
Tech Fellow
CodePath · Jun 2026 – Aug 2026
AI Automation Extern
Outamation · Dec 2025 – Feb 2026
Education
City University of New York, Queens College
Bachelor of Arts in Computer Science · Class of 2026
Skills
✅ Verified by Extern
An Outamation externship on AI-powered document intelligence, focused on turning unstructured mortgage files into searchable data through OCR, extraction, and RAG pipelines.
The externship work assembled an end-to-end document intelligence pipeline for mortgage files, combining OCR, extraction heuristics, and retrieval-augmented generation. Deliverables include Colab notebooks, a Gradio chatbot tied to a local RAG pipeline, OCR comparison results, and a slide-deck report summarizing evaluation findings.
I produced a runnable, end-to-end AI document pipeline with extraction code, OCR comparisons, RAG notebooks, a Gradio chatbot prototype, and a slide-deck that documents accuracy and evaluation findings.
Built an end-to-end fraud detection pipeline on 1.3M+ transactions, training on Random Forest, XGBoost, and Logistic Regression to achieve a 99.53% accuracy and 0.97 ROC-AUC. Mitigated class imbalance using SMOTE and identified 15 high-impact features via SHAP analysis, boosting precision from 45.23% to 57.89% while improving model explainability. Collaborated with a team of 4 to optimize classification threshold and conducted 10-fold cross-validation to boost recall to 66.76% on a heavily imbalanced dataset.
Achieved 80.05% test accuracy in classifying skin lesions by developing a Skin Cancer Prediction tool using Convolutional Neural Networks (CNN) and Logistic Regression on the HAM10000 dataset. Enhanced skin classification accuracy by 17% by implementing advanced architectures like ResNet-50 using TensorFlow, resulting in a 97.50% test accuracy. Reduced model bias by implementing data augmentation techniques and preprocessing steps using normalization and class balancing, visualizing model performance using confusion matrices in Matplotlib.
Trained a mathematical model to predict the amount of calories burned during exercise among 15,000+ individuals using XGBRegressor. Utilized NumPy and Pandas to handle and manipulate datasets effectively, performing tasks such as data wrangling and outlier detection. Scripted the predictive calories burnt model and obtained a prediction accuracy of 90.04%.