Jamila Rejeb

Deploying new ML project to the group brands enabling them to better understand their clients and improve their sellout. Utilisation des LLM afin de synthétiser des articles de veille dans le domaine de la beauté Keywords : product recommendations, client segmentation, python, Dataiku, GCP, LLM, prompt engineering

▌ Product chaining - Computing product similarities in order to link them and maintain coherency in terms of price changes ★ Keywords: Vertex AI, transformers, GCP, SQL, LLMs ▌ Pricing Elasticity - Modeling pricing elasticity taking into account high inflation context ★ Keywords: Airflow, GCP, Docker, time series, SQL

▌ Malware detection based on browsing logs - Distributed pipeline of preprocessing (Spark) - Very small amount of malware data - Feature engineering using cybersecurity experts' knowledge - Creating CICD pipeline ★ Keywords: Spark, Anomaly detection, Oozie, Ansible, Jenkins, ▌ Clear password detection - Importance of detecting all the rows with potential passwords - Only 0.2% of the dataset contains real passwords - Predictions interpretability - Deployment of the solution in production - Data shift and model performance monitoring with Mlflow ★ Keywords: NLP, Reinforcement Learning, Active learning, word embedding, recurrent neural network, MLFlow ▌ logs scoring - collecting logs from servers - Data processing - scoring logs - only alerting users when a relevant log occurs - compute logs similarities to prevent sending the very similar alerts ★ Keywords: NLP, druid, kafka, bash, Lime

▌ Preprocessing and classifying a large dataset of patents (500 GB + in textual relational tables). Each patent is linked to one or multiple IPC indexes (International Patent Classification) in a parallel fashion in a cloud environment. ★ keywords: Big data, Google Cloud Platform (BigQuery, Compute Engines, Google storage), Apache Parquet, Dask, NLP, Hierarchical classification

▌ Computer Vision: Automating accident report form analysis using image processing (SIFT, HOG) and deep learning (Tensorflow API, YOLO). ▌Natural Language Processing: Used TFIDF and distance metrics to compare PDF generated files with their source files to detect possible errors. ★ Keywords: Python, DATAIKU, OpenCV, SIFT, HOG, Tensorflow API

Developing Proof of Conceptions as part of Fin'IT team at Societe Generale Corporate and Investment Banking - SGCIB. ▌ TIPS: a tool designed to foresee valuations of the underlying price based on a chartist approach/. - Collecting previous prices -Cleaning data, filtering Out Market Noise using Zig-Zag Indicator -Creating a database of possible patterns and their outcome as a training data for the next phase -Using Scikit-Learn machine learning algorithms to predict the evolution of the price curve ▌ Data crunching for MIFID 2 -creation and maintenance of a database of registered financial derivatives prices -working with senior traders on adjusting the pricing strategy based on previously registered prices -Implementing the algorithm using python: Using MangoDB swaggers Data crunching in order to prepare data Executing VBA macros through Python and saving the price curves for each derivative. ★ Keywords: Python, Pandas, Numpy, Scikit-Learn

Data analysis for a smart wifi project: SmartAP ★ Keywords: Java, Highcharts.