Saber Hemaidia

I spearhead the design and industrialization of Generative AI solutions for wealth management. By bridging the gap between complex AI architectures and business requirements, I deliver high-performance, compliant tools that directly enhance operational efficiency. • Engineered and deployed secure, GPT-powered solutions via Azure AI Foundry, leveraging advanced RAG pipelines and autonomous agents (LangChain, LangGraph) to automate wealth management workflows. • Architected highly scalable, production-grade backends using FastAPI. Established robust LLM observability and telemetry (Langfuse, MLflow). • Directed cross-functional prompt engineering workshops, collaborating with hundreds end-users to translate operational bottlenecks into high-value AI use cases, improving model output relevance and adoption. • Navigated rigorous regulatory and security constraints within an international banking environment, partnering seamlessly with IT and compliance teams to ensure agile, risk-free AI deployments.

Due to the exceptional circumstances of the energy crisis, Enedis created the OPH team to collect and structure citizens' electrical consumption data to conduct studies on energy-saving efforts. For example, we managed to save the equivalent of the electricity consumption of a city like Paris by cutting off water tank electricity during peak hours. • Analyzed electric consumption patterns across various sectors in France (OTE). • Scheduled and managed the transmission of consumption data to the Open Data website. • Developed a forecasting model for electricity consumption across different segments. • Created intuitive data visualizations using PowerBI. • Engineered packages and documentation for efficient extraction of electric consumption data from Enedis's servers.

To enhance user experience on a website, I developed a computer vision API that convert user faces into personalized avatars. For this mission I had to find multiple datasets of image face. Train multiple CNN models using AWS s3 to classify the different part of the face to generate the corresponding avatar. • Led the implementation of a computer vision API to generate personalized avatars from user photos. • Designed and trained Convolutional Neural Network (CNN) models for accurate facial feature classification. • Employed U-Net based semantic segmentation to precisely extract facial hair and hairstyles from user images. • Collected and meticulously labeled thousands of facial images to ensure high-quality training data. • Conducted extensive image pre-processing to optimize data for model training. • Performed Principal Component Analysis (PCA) and K-Nearest Neighbors (KNN) classification for feature dimensionality reduction and robust classification. • Leveraged AWS infrastructure for efficient training of neural network models, ensuring scalability and performance. • Successfully deployed the neural network models as a user-friendly web application using Django, providing seamless integration with the website.

Altalia Courtage an insurance company needed someone to clean the data of their customers to make it more readable. I've then convinced them of the utility of Machine Learning to assist them with their commercial proposals. - Data cleaning and reorganization of personal user data - Analysis of customer data. - Realization of graphics and documentation. - Implementation of a Random Forest model giving as an output a price range for commercial proposals. - Deployement of a data dashboard for the commercials using Django Technical environment: Python, Sci-kit, Pandas, Numpy, Django, Matplotlib, Excel

Replacing a numerical simulation of the reflectance of an optical surface with Deep Neural Network model taking as inputs the geometry shape and its physical properties. - That output a good approximation of the optical reflectance. - Deployment of a Machine Learning model (Radom Forest) and CNN model with Flask - Technical documentation. Technical Environment: Python, Sci-kit, Pandas, Numpy, Keras, Matplotlib, Django

Various missions on different Renault autonomous driving car prototypes as an ADAS and camera specialist. - Development of driving assistance scripts using camera, Lidar and ECU data. Test and Validation of camera(Sony) and Soft Computer Vison (Mobileye) suppliers. Technical documentation and presentation. - Data collection and fusion of the Lidar, camera and ECU for 3D-mapping of the environment around the car. - Camera calibration. Technical environment: Python, C++, Matlab Sci-kit, Pandas, Numpy, OpenCV, Tensorflow, Keras, ROS, Signal processing

Context: 6-month Master internship to develop the business 'from scratch' of an add-on for confocal microscope intended for researchers in biology and physics. Duties: Evaluation of potential users of the add-on and demonstration of the use of the add-on in a conference or in the laboratory. Marketing study Canvassing by email and telephone Demonstration of the add-on during QBI 2019(Rennes) and FOM2019(London) microscopy exhibitions Customer visit for demonstration of commercial proposal. Technical Environment: Marketing Study, Prospecting, Matlab, Optical Instrumentation, Technical Documentation…

Collecting data from scientific publication to prepare a numerical simulation on C++ of the different plasma reactions happening in an ion thruster rocket used for NASA satellite.

During that intership of 2 months, I worked with Antonio Monari as my referant, I've learned how to use VMD software and to program ab initio simulation with a QM/MM method using Gaussian, VMD software and Amber mechanical force fields program. Subject: Biophysical study of DNA degradation, correlation between structure and reparation efficiency. Task: Bioiformatics, Numerical simulation, research of scientific publication in english, programming, uses of Linux and bash language.

An internship of one month with Oudich Mourad as my referant. We studied together the simulation of an surface acoustic wave on a piezoelectric device with Comsol multiphysics. Subject: Acoustic wave surface Comsols multyphysics’s simulation of a piezoelectric device.