Mohammad Saghali

Academic Advisor: Prof. Luca Vassio · Master's Thesis conducted in collaboration with Nokia Bell Labs

• Implemented, optimized, and tuned end-to-end AI/ML pipelines for the productization phase, leading core software engineering, module integration, and performance benchmarking.
• Developed a product-bound, full-stack ML Intrusion Detection System combining low-overhead PCAP preprocessing, temporal feature extraction, and payload analysis to facilitate real-time, cross-environment AI decision-making for high-throughput data streams.
• Designed a scalable architecture for online model training planned for productization by Nokia Cloud and Network Services (CNS).
• Co-authoring upcoming publications on AI/ML and clustering techniques, including a journal article and an academic book chapter.

• Developed and implemented an ML-based Intrusion Detection System (IDS) for network security, showcased at the Bell Labs Centennial and now integrated into Nokia's CNS productization roadmap.
• Implemented and co-designed DASNet (Deception and Adversarial Shield Network), a novel GAN-based defense architecture to protect AI-enabled RAN features; generated adversarial samples and executed stress testing against evasion attacks (presented to Mobile Networks in ongoing FSY).
• Co-developed ACENet (Adaptive Clustering and Embedding Network), an unsupervised clustering and embedding method for robust, adaptive anomaly detection in network streams.
• Architected and built a complete real-time AI decision-support visualization dashboard developed with Dash and Plotly, showcased to global stakeholders at Mobile World Congress (MWC) 2026, RSA Conference 2025–2026 (San Francisco), Core User Group (CUG) 2025–2026, and customer workshops.
• Curated and preprocessed large-scale network traffic datasets to benchmark ML models, resulting in two filed Nokia patent applications and an IFIP Networking 2025 publication.

• Engineered a reinforcement learning-based neural network utilizing autonomous agent neurons to execute complex decision-making and classification tasks. Extended this foundational thesis research using Python, PyTorch, Scikit-Learn, NumPy, and Matplotlib, culminating in an academic paper published in the Journal of Theoretical Biology.

• Developed web applications end to end with Angular, HTML5, CSS3, and TypeScript, collaborating with product teams using Git through the full development cycle.

Recognized by supervisor at Nokia Bell Labs for outstanding contributions.

Politecnico di Torino grant funding the research internship.

Merit-based funding for thesis research abroad.