Self-supervised detection and explanation of anomalies in time series data for large-scale network security. Research focus on developing robust deep learning architectures for real-time anomaly characterization, and LLMs combined with RAG-based methods for automated interpretation using business knowledge bases, eliminating human expert intervention.

Research on self-supervised deep learning for time series anomaly detection and explanation in collaboration with Orange Innovation. Developing novel XAI approaches for automated anomaly characterization and LLM-enhanced interpretation frameworks. Working towards publications in ML conferences.

Unsupervised anomaly detection system for heterogeneous data (logs, metrics, traces) using pattern variation detection and time series correlation. Optimized real-time pipeline reducing memory by 70% and improving scalability 3x for 10M+ events/day. Built RAG-enhanced LLM agent for automated anomaly reports with root cause analysis.

Time series trends analysis using tools such as Prophet, SARIMA, STL, and DTW. Identified correlations between logs and metrics to predict system behaviors. Built a robust pipeline for live data retrieval from internal APIs with integrated data preprocessing.

Developed a LoRaWAN-based emergency messaging system, covering 25km range, ensuring emergency communication in remote areas without internet access.

Gained hands-on experience in full-stack development, alongside senior developers.

Introduced to processor architecture, with a Senior SoC (System-on-chip) Designer.