Anas Zilali

- Developed a RAG chatbot framework specialized in multiple domains. The chatbots integrate source citation functionality and provide precise answers derived from document analysis, surpassing the capabilities of standalone LLM. - Developed NexiaSynth, an application to synthesize and anonymize tabular data using Deep Learning models (GAN, VAE).

- Developed an automated ETL using Dataiku for monthly management control reports on insurance products, significantly improving operational efficiency and reducing operational risk. - Implemented business rules to verify the calculations produced by the Dataiku flow, ensuring the integrity of the results.

Develop and deploy text recognition (OCR) models. The resulting models are used in Cardif's pipeline for extracting information from scanned documents. The aim of the pipeline is to automate and accelerate the processing of claims documents in all geographies (multilingual approach).

Develop time series forecasting models based on statistical and Machine Learning models. The resulting models are deployed to provide industrial insight in factories based on available data. - Preprocess and generate features from big data (more than 1M row). - Implement and train recurrent models RNN/LSTM/GRU/Conv1D... - Implement statistical models ARIMA/VAR/Cointegration... - Detect anomalies and predict industrial key performance indicators.