Deniz Beyazgul

Location: Manchester, UK (Hybrid)

Project: VISTA – Visual Intelligent System for Threat Analysis

• As a Data Scientist on the VISTA project, I manage a multidisciplinary team of five, driving research and development of intelligent systems for threat analysis in industrial IoT environments. My work focuses on integrating and modeling complex multi-source data for predictive analytics and explainability.

• Conducted extensive research into data integration techniques for heterogeneous industrial IoT datasets, focusing on mid-fusion strategies.

• Developed an unsupervised deep learning framework combining autoencoders and manifold alignment to integrate disparate datasets and extract meaningful latent representations.

• Designed and trained multi-class classification models using Random Forests for threat prediction, optimizing for performance and robustness.

• Implemented explainable AI (XAI) methods, including Global SHAP, Local SHAP, and LIME, to provide interpretable insights into model behavior and decision pathways.

• Applied the same methodology to the publicly available WADi (Water Distribution) dataset (a complex multisensor dataset), demonstrating the generalizability of the threat detection pipeline across different domains.

• PPC Campaign planning and management

• Conversion & Sales Tracking

• KPI Reporting: Campaigns Metrics & Creative Performance

• Client Data Analysis & Visualization

• 12.5 X ROAS with only $20K ad spend in 4 months

• PPC Campaign planning and management

• Conversion & Sales Tracking

• KPI Reporting: Campaigns Metrics & Creative Performance

• Client Data Analysis & Visualization

• 2 million USD in ad spend