Mohamed DYAB

Deployed data pipelines for the Racetrack Agent, an agentic AI procurement tool that automates complex approvals and policy checks. Engineered data foundations using Snowflake, dbt, and Airflow to feed context into the AI agent's decision-making process.

Developed a standard template for integrating data quality checks on dbt and Airflow pipelines that is used by at least three different teams. Provided expertise for the Global Data Operations teams to adopt data quality best practices.

Voluntary mentor for learners enrolled in the Data Engineering Professional Certificate, supporting their progress after my successful completion of the program.

Architected, designed, and implemented a data lake and data quality pipeline on an AWS account from scratch after a successful Proof of Concept for the Global Data Operations team. Created all the IaC and CI/CD scripts as well as ETL pipelines, as well as integration with Redshift, Power BI, and other AWS accounts, as well as API calls to execute the data quality checks. Implemented Data Quality KPIs for Logistics, Procurement, and HR in collaboration with different business units. As a Tech lead: Led a POC using dbt to propose dimensional data modeling as an alternative data model to Data Vault modeling for a Healthcare business unit. Designed and validated a custom data quality proof of concept, subsequently adopted by Air Liquide, resulting in annual 5-figure cost savings compared to existing commercial solutions. Executed a comprehensive demonstration of the data quality pipeline on AWS and visualized Data Quality KPIs on Power BI.

Migration from enterprise data warehouse towards a single unified cloud data platform (AWS S3, Glue, Snowflake). Developing and maintaining data pipelines. Creating data models and iterating with different stakeholders.

Streamlined and documented the process of ingesting new business events to the data platform and exposing them to Snowflake for various data squads. Onboarded new squad data engineers regarding business events ingestion. Migrated several business events to Snowflake from a newly set up secure Kafka cluster as a part of a data security initiative. Engaged in a comprehensive audit of Snowflake's usage and cost structure, identifying optimization opportunities to reduce expenditure.

Extensive hands-on entrepreneurship learning experience: co-founder selection, ideation, and validating customer needs/market. Our team pitched for the pre-seed investment committee.

Improved the accuracy of conversion rate prediction of Google advertising campaigns by taking advantage of seasonality using LSTMs.

Talking face detection in videos: showed that using an LSTM to combine visual and audio modalities performs better than using either modality alone. Used 400 GB of video to train CNNs and LSTMs to detect talking faces in videos using audio-visual features. GitHub: https://github.com/dyabmo/plumcot

Co-authored "A Comparison of GPU Execution Time Prediction using Machine Learning and Analytical Modeling" at the IEEE-NCA 2016 conference. Trained several machine learning models to predict the execution time of applications running on GPUs and compared them with an analytical model. GitHub: https://github.com/marcosamaris/gpuperfpredict

Successfully delivered embedded software components of the eCall system (Emergency response to car accidents) and successfully​ tested​ them on hardware boards.