Omar Abdesslem

Power Grid Load Forecasting using Machine Learning

A data-driven thesis modeling Swiss electricity load using Time Series and Machine Learning. Forecasts were informed by seasonality and enriched with exogenous weather data. The project applied rigorous time series diagnostics and evaluation metrics.

AI-based Log Anomaly / Cyber-attack Detection for SOCs

During half of my summer internship, I customized and deployed an inhouse trained LLM for real-time Security log analysis. Modified prompts and output normalization; added Phi-4 severity ratings; built Streamlit dashboard for real-time monitoring. Achieved up to 99.98% accuracy; end-to-end inference on 10k logs in ~2 minutes with minimal resource overhead; near real-time processing on a single GPU.

SwissGrid: Prediction of PV installation angles

Winner team for the Swissgrid Challenge for Energy Data Hackdays, Aargau.

For this project, we used Databricks and Python to predict installation angles for 233k+ PV plants in Switzerland. We applied geospatial clustering and predictive modeling to estimate missing data and improve renewable energy forecasts.