Refactored current code and added tests, and increased test coverage from 42% to 80%.

For over 20 Sklearn, XGBoost and LightGBM machine learning models supported, implemented automatic hyperparameter value verification and incorporated into both the WebApp and DS library. Used Python and Jython.

Created new admin pages to monitor various ML build and resource usage statistics. Implemented backend logics, controller end points and admin authentication. Used React, Redux and Java.

Built prototype of Alexa extension for Indeed, and won 2nd Place and ‘Most Likely to Succeed’ award in hackathon

Designed Python algorithms to tokenize 3 million+ Arabic text data, e.g. sentence segmentation and POS tagging

Implemented Naïve Bayes classifier and Turney Algorithm to perform Arabic NLP tasks including sentiment lexicon bootstrap, feature extraction, Tf-idf marking and sentence polarity analysis

Created Arabic Sentiment Analysis Models (linguistic and statistical patterns), as core of relevance prediction

Conducted evaluation and benchmarking of software alpha version, improved recognition accuracy by 67%