Ahmed Magdy Abdelazeem

Cairo, Egypt +20 -- Ahmed Magdy Machine Learning Engineer with +3 years of experience in machine learning, and natural language processing, have hands-on experience in Python and machine learning libraries including Sklearn, PyTorch, Huggingface. EXPERIENCE AIFund, Medellin, Colombia — Machine Learning Engineer Oct 2019 - PRESENT Rustle.co ● Developed an automated data pipeline for extracting/scraping data from CS websites using MongoDB and S3. ● Implemented a Question-Answering NLP system over HTML text to answer CS questions given a company knowledge base using AWS and Fast-API. ● Deployed a Search Engine system using Elasticsearch. ● Developed a silence detection model to measure the quality of CS calls achieving 89% accuracy. Factored.AI ● Implemented a Zero-Shot classification system for various use cases using NLP SOTA, dockers, and Fast-API. ● Created an HR matching tool, which helps in matching candidates with job descriptions and gives a score to each candidate, using various NLP algorithms (Bert, Rake, TF-IDF). ● Implemented ESA model to encode job titles as concepts and measure similarities achieving >85% accuracy. INTDV, Cairo, Egypt — Machine Learning Engineer Nov 2018 - Apr 2018 ● ● Developed Arabic NLP modules like chatbots, intent detection, and NER. Developed an ML model to detect duplicates in a large database. ● R&D in stock market prediction with reinforcement learning and time series forecasting. ● Developed an automated testing system for stock market predictions using trading indicators and hypothesis testing. EGYx.ai, Cairo, Egypt — Machine Learning Engineer May 2017 - Aug 2018 ● Developed Arabic Chatbots for restaurants' customer service. ● R&D in AI Arabic home assistant. Projects Novel Ensemble Learning Approach to Unsupervised Record Linkage ● Implemented a semi-supervised approach to solve the record linkage database problem. ● The proposed approach is to train with self-learning and adding automatic seed selection. ● Achieving around 85% accuracy outperforming simpler methods like TF-IDF. Tweets Author verification ● Research project in which I try to outperform some of the benchmarks with simpler methods. ● Calculating Z scores for top N-words and compare them together. ● Used SVM classifier with a different set of features like lexical diversity and number of hashtags. ● Used one-shot learning with the Siamese network and achieving 87% accuracy. MASRI(GP) ● An Arabic home assistant that takes voice commands and performs services. ● Implemented the Voice recognition modules, trigger word, Text To Speech. ● The project won IEEE Hackathon and joined the Nile University accelerator program and got funded. EDUCATION Helwan University, Cairo, Egypt — Bachelor’s of Engineering Sep 2013 - Aug 2019 Major: Computer Engineering GPA: very good (the tenth)