Projects

ESSEC Metalab • Dec. 2024 - Present • Cergy, France

AI Projects Leadership:
- Leading all AI and research-related projects, overseeing development and implementation of cutting-edge AI solutions.
Team Management:
- Managing a team of professionals to drive innovation within the lab, fostering collaboration and technical excellence.
Research Initiatives:
- Initiating and overseeing research projects in AI, contributing to academic publications and industry advancements.
Technology Stack (Python, TensorFlow, PyTorch, LLMs, DevOps):
- Utilizing advanced AI frameworks and tools to develop innovative solutions, integrating best practices in DevOps and MLOps.

AZ+ • Jan. 2024 - Present

AZ+: SuperApp Architect and Management (Microservices):
- Led a team of 13 in the design and deployment of the SuperApp AZ+ for Madagascar, offering services such as e-commerce, food delivery, and ride-hailing under one platform.
Cloud Migration (GCP, AWS, Kubernetes):
- Executed the migration of infrastructure from Firebase to AWS and GCP, improving scalability, reducing costs, and increasing control over the deployment process; used Kubernetes for managing containerized apps.
Self-hosted Services (PostgreSQL, GitLab, Minio S3, Grafana):
- Set up and managed self-hosted services on internal servers, including GitLab, Minio S3 (object storage), and Grafana (monitoring and analytics), ensuring full ownership and cost control over the infrastructure.
Technical Innovations for Local Solutions:
- Developed innovative location tracking solutions specifically tailored for Madagascar, enabling more accurate delivery management and solving challenges not addressed by existing tools in the region.

SERMA Safety & Security • Feb. 2024 - Aug. 2024 • Rennes, France

PCB Component Recognition (YOLO, OCR, Fuzzy Matching, OpenCV):
- Developed a system to recognize Printed Circuit Board (PCB) components using YOLO (object detection), OCR, and Fuzzy Matching techniques, reducing component misidentification by 35%.
Reverse Engineering with LLM4Decompile (LLM4Decompile, Assembly Code):
- Integrated LLM4Decompile to automate the reverse engineering of binary files from detected hardware components, streamlining the process of extracting and analyzing firmware from embedded chips.
Machine Clustering and CI/CD Automation (GitLabCI, MLFlow, Ansible, K8s):
- Clustered a set of 5 spare machines with Kubernetes and implemented fully automated CI/CD pipelines using GitLabCI and MLFlow, automating the deployment and monitoring of ML models.
Graph RAG & LLM POC (GraphRAG, Neo4J, Llama3):
- Developed a POC that uses GraphRAG for enhanced document retrieval and analysis, leveraging self-hosted Llama3 and Neo4J for local document integration, running on local machines.

IMT Atlantique • May 2023 - Sep. 2023 • Rennes, France

Data Acquisition & Processing (ETL, PowerBI, Pandas, Seaborn, Python):
- Acquired and processed data (ETL with PowerBI) from ANFR to understand the power usage of every antenna (Base Stations) in France since 2015 using Pandas, Seaborn, and Pyplot.
Energy-Saving Strategies (Seaborn, Pandas, Python):
- Proposed four strategies to save energy for mobile networks based on key factors, improving the existing models.
Real-World Application (PowerBI, Python):
- Implemented a GenAI Model (CTGAN) to synthesize the traffic data of base stations in Rennes, resulting in energy savings of up to 20%.

Videotron • Sep. 2020 - Jun. 2022 • Cairo, Egypt

Customer Satisfaction Improvement (Excel, SQL):
- Developed data-driven strategies that increased customer satisfaction from 80% to 95% over a one-year period by identifying key factors contributing to dissatisfaction and addressing them through targeted interventions.
KPI Analysis & Agent Performance (PowerBI, SQL):
- Conducted in-depth analysis of agent performance, resulting in a 20% improvement in key performance indicators (KPIs) by identifying areas for training and skill development.
Process Optimization (PowerBI, Excel, SQL):
- Streamlined resource management and optimized workflows by analyzing callback data and process inefficiencies, enhancing decision-making and improving response times within a seven-day period.