Ahmed's Website

About

Learn more about me

Data Scientist

My journey started in the suburbs of Saida, an western town in Algeria. My childhood was pretty much similar to any other kid in the area; playing and school. In high school, I was, for the first time, introduced to a computer. I loved it and decided to pursue my studies in learning about this magical box! I got a bachelor's and a master's degrees in computer science. Then, I got a scholarship to pursue a PhD at Beihang University in China.

My research focuses on semantic web, knowledge graphs, machine learning, deep learning, and NLP for Arabic. Besides, I enjoy coding in Python and Java, building predictive models, and learning new tech.

On the side, I like reading religious books, playing football, recording recitations, and having fruitful debates.

Website: www.ktobah.com
City: Saida, Algeria

Phone: +213 667334948
Email: ktobah@gmail.com

Projects

Trained Models

GPU Hours

Certificates & Awards

Skills

Python 90%

Machine Learning / Deep Learning 80%

Tensorflow / SK-Learn80%

Research / Knowledge Graph 90%

Numpy / Pandas / Seaborn 90%

Linux / APIs / Docker / AWS 80%

Interests

Research

Reading

Intellectual Discussion

Tech & Coding

Open Source

Fundraising

Football

Traveling

Testimonials

Ahmed is a talented person with a rare combination of patience, bravery, and research skills. He and I were colleagues at Beihang University. He was always a good team player, determined and a hard worker. He's great in computer science.

Boualam Belaidi

Automation Engineer

It's been a pleasure working with Ahmed at ControlExpert, where we've done a lot of projects together. As a data scientist, I admire his excellent professional knowledge and hardworking quality. He has brought a qualitative leap forward to our team's modeling ability.

Zhi Pan Li

Data Scientist

One of the ambitious and creative people that I ever met.

Mohamed Reda Soltani

Researcher

Resume

Check My Resume

Sumary

Ahmed Ktob

Detail-oriented professional with 2+ years of experience and proven knowledge of machine learning, data analysis, and predictive modeling.

92 El-Kadissia, Ain El-Hadjar, Saida 20001, Algeria
(+213) 6-67334948
ktobah@gmail.com

Education

PhD of Computer Software & Theory

Sep 2014 - Jun 2018

Beihang University, Beijing, China

My research focused on the Arabic knowledge graph which led to 4 publications. This opportunity allowed me to strengthen my research skills and learn how to efficiently collaborate.

Exchange Program in Computer Science

Jan 2018 - May 2018

University of Liège, Liège, Belgium

This program allowed me to discover how higher education works in Europe. I took classes in advanced machine learning, robotics, and bioinformatics.

Certificate of Chinese Language

Sep 2013 - Jun 2014

Beihang University, Beijing, China

I was introduced to the Chinese culture through language learning. I had courses in reading, listening, speaking, and writing. My overall grade was an A.

Master of Networks, Systems and Information Security

Sep 2011 - Jul 2013

Djillali Liabes University, Sidi Bel Abbès, Algeria

I learned a great deal about the different aspects of security that touches on networks, databases, and operating systems. My research thesis was about extracting structured data from HTML pages and converting it to RDF.

Bachelor of Computer Science

Sep 2008 - Jul 2011

Dr. Tahar Moulay University, Saida, Algeria

My first intro to computers. I took many introductory and advanced courses in math and computer science. I ended up writing my first thesis about sniffers in a local network.

Professional Experience

Data Scientist

2019 - Present

ControlExpert, Beijing, China

Identify valuable data sources and automate collection processes.
Analyze large amounts of information to discover trends and patterns.
Undertake the pre-processing of structured and unstructured data.
Build car damage detection, segmentation, and decision aggregation models using ML/DL techniques.
Explore structured data & implementation of NLP algorithms.
Propose solutions and strategies to business challenges.
Perform extensive hyper-parameters tuning.
Combine models through ensemble modeling.
Develop auto labeling tools.
Report findings through data visualization techniques.
Collaborate with engineering and product development teams.

Researcher

Sep 2014 - Jun 2018

Beihang University, Beijing, China

I worked at Prof. Zhoujun Li's lab as a researcher on semantic web and knowledge graphs. My research project was on the construction of an Arabic Knowledge Graph. I identified the challenges facing its creation and the opportunities it offers. We were able to create a small knowledge graph and built two tools that would allow practitioners and data owners to transform their data into RDF and share it with the world. The applications of the AKG are beyond description; it can be used in creating semantic-based Arabic search engines, build smarter Arabic AI applications, and much more.

Software Test Engineer

Oct 2015 - Apr 2016

Symbio, Beijing, China

Design test plans, scenarios, scripts, or procedures.
Document software defects, using a bug tracking system, and report defects to software developers.
Identify, analyze, and document problems with program function, output, UI, or content.
Monitor bug resolution efforts and track successes (using Jira).
Provide feedback and recommendations to developers on software usability and functionality.
Conduct software compatibility tests with hardware and operating systems.

Full Resume

Download my full resume

Projects & Publications

My Projects

Structured Data & Graph DB

A highly-collaborative project that aims to enhance CE's benefit from the gathered structured data. I work with the team in analyzing the data and extracting a data model to build a graph database using Neo4j. This database will allow CE to explore the hidden connections between claims, car drivers, reporters, investigators, and so on to eventually discover frauds.

Damage & Parts Segmentation

This project's goal is to solve an issue with part detection that prevents us from delivering accurate results on the damaged parts. Also, we aim to quantify the damage so I started working on both car damage and parts segmentation. The initial results seem promising.

Damage Detection

Lead the development of a brand-new model to detect the damaged parts in a car accident. This predictive model will be integrated into our risk control service to improve our accuracy. The first step to achieving such a goal is data. Thus, I collect data, check it for completeness, clean it, upload images to our AWS cloud, and update the labeling MongoDB database.

Aggregation & Thresholds

Technically, decisions are taken around the nuances of thresholds. Therefore, a well-defined threshold can make a significant impact. In this project, I continuously work on optimizing our thresholds for better decision making. Since a decision that is taken from several images is better than from a single one, Tensorflow and Optuna were used to build an aggregation model and run an extensive hyper-parameters search to find the optimal values ( 10-15% accuracy boost).

AutoLabeler

At CE, we recognized the bottleneck of slow data labeling. Thus, I worked on creating a labeling tool that used one of our internal object detection models to automatically label our images. Further, human labelers would check if the results are accurate and amend them if necessary. This tool had an accuracy of 80% and allowed us to, approximately, save over 400 hours of human labor.

MongoDB Management

Our labeling tool uses a MongoDB database to store data. I am responsible for keeping the database up to date, add and remove data, fix any issues that may arise, and automate the generation of labeling stats for our human experts. Also, I create scripts that are used by the team to generate data for model training. In addition, I was one of two members responsible for checking the quality of our human experts until August 2019.

My Publications