nouha jegham

*Apply parameter-efficient fine-tuning, incorporating QLora and prompt instructions to tailor Microsoft’s SLM Phi-2 for a Sentiment Analysis task. *Finetune MT5-Small model for translation tasks from English to French and from English to Arabic. Implement 4-bit quantization to optimize computational efficiency. *Develop a FastAPI for model integration and deployed it on Azure cloud, ensuring custom domain integration and SSL/TLS encryption via Let’s Encrypt certificate for secure communication over HTTPS.

*Gather hotel data from online sources by employing Beautiful Soup for web scraping. *Use of Random Forest Regressor, achieving an impressive R-squared value of 99%, accurately predicting total energy consumption for each region in France. Implement Named Entity Recognition (NER) to identify potentially mislabeled disease names within doctors’ prescriptions. * Visualizing Data using Looker Studio and PowerBI

*Used signal processing and statistical techniques like kurtosis, envelope spectrum and Kurtogram, to analyze vibration data of the gearbox. Enabling for the extraction of essential features, providing insights into the machinery’s condition. Effectively clustered the data to differentiate between operational modes of the gearbox using vibration data, laying the groundwork for failure detection step. Utilized kurtograms to identify failure indicators and generated them for the strong vibration mode.