03 / Arabic NLP research & services
Lableb
Aug 2020 — Sep 2023AI Engineer · Remote
Working with the complexity of Arabic.
At Lableb, I worked across Arabic NLP research and service engineering. I joined as an AI engineering intern from April to July 2020, before moving into the full-time role in August 2020.
Research & engineering
Researching better spelling correction
I compared n-gram and neural noisy-channel approaches through iterative spelling-correction experiments. KenLM improved runtime while giving comparable results; I versioned related spelling artifacts with DVC.
Model research
Training and evaluating Arabic models
I researched, trained, and evaluated named entity recognition, part-of-speech tagging, transliteration, and spelling-correction models in Azure Machine Learning. Evaluation included task-appropriate measures such as F1, precision, recall, and accuracy.
Separate production outcome
A spam service in production
I built a B2B spam-classification service that handled up to 30,000 requests per day and filtered about 90% of spam queries.
My service engineering work included NLP microservices built with Python, Go, Flask, Spring Boot, and Go Fiber.