Rachid Reda DOKKAR
Data consultant at DXC Technology and Data Science graduate from the University Paris Cité. Computer Science engineer from École Nationale Supérieure d'Informatique ESI Alger with 12 months of professional experience in computer vision and machine learning. Seasoned participant in machine learning competitions on Zindi (https://zindi.africa/users/ridadokkar), specializing in data engineering and data science consulting.
Azure AZ-900 and AI-900 cloud certifications, Top 25% in my first school ESI.
View My LinkedIn Profile
View the Project on GitHub DokkarRachidReda/portfolio-data-science
Selected projects in Data Engineering
Krello Data Engineering on Azure
Developed end-to-end data integration solutions using Azure PostgreSQL, Azure Data Factory, and Snowflake, seamlessly transferring private dev data from Krello (a startup I co-founded). Implemented ETL pipelines for PostgreSQL to Snowflake, orchestrated web scraping workflows with Apache Airflow, and automated data lake-to-warehouse integration.
*The databricks part has not been implemented yet.
Azure, Azure Data Factory, Azure Database for PostgreSQL, Snowflake, Apache Airflow, Databricks, SQL, Spark, ETL/ELT, Batch Jobs
Selected projects in machine learning, Computer Vision, and NLP
Zindi New User Engagement Prediction Challenge
The objective of this challenge is to predict if a new user from a selected cohort will engage with Zindi in their second month on the platform. For example, if a user joined in June, will they be active in July?
I have proposed a model based on a deep auto-encoder and a classification network trained jointly to learn the features that are important to model, then use them for the classification.
Ranked 5/212 !
Neural Networks,Auto-encoders, Variational Auto-encoder, PCA Analysis, Visualisation, Pytorch, Pandas, Numpy, Python
Cain’s Jawbone Murder Mystery
My job, as the puzzle solver, was to try to put 75 pages of the famous Cain’s Jawbone into its correct order, using AI NLP algorithms.
I have used BERT NextSentence prediction (NSP) from Hugging Face along with cosine similarity to order the pages.
LSTM,RNN, BERT, Transformer, NLP, Pytorch, Pandas, Numpy, Python
NASA Harvest Field Boundary Detection Challenge
The ultimate challenge: crafting a cutting-edge segmentation algorithm for crop field boundaries detection.
I have utilized a combination of VGG16 and Unet++ to propose an advanced segmentation algorithm for crop field boundaries detection.
CNN,VGG16, Unet, Transformer, Vision Transformer, Computer Vision, Pytorch, Python
Video Classification: Human Activity Recognition
Propose a new architecture for the human activity recognition using Deep Learning.
I have utilized a combination of 3D-CNN and Vision Transformers to propose an advanced video classification algorithm. The proposed algorithme has achieved SOTA results on 3 public datasets.
3D-CNN, Transformer, Vision Transformer, Computer Vision, AI Explianability, Pytorch, Python, Virtual Machines
