Achieved 99% accuracy in detecting network threats with an ensemble AI framework, fortifying cybersecurity defenses in real time.

Designed and implemented a novel ensemble model combining supervised (Random Forest, XGBoost) and unsupervised (Autoencoder, Isolation Forest) learning techniques to detect both known and unknown network threats with 99% accuracy, leveraging cutting-edge traffic data.

Processed and optimized over 2.5 million traffic instances, reducing dataset size by 60% through feature selection and cleaning, ensuring efficient model training and accurate predictions.

Developed a meta-learning framework that integrated base model predictions and anomaly scores into meta-models (Logistic Regression, Random Forest), achieving high validation accuracies of over 99% and superior F1 scores. This approach excelled in handling imbalanced datasets, enhancing detection of rare and critical events.

Impact: Delivered a robust real-time intrusion detection system that enabled proactive defense against evolving cyber threats, significantly improving network security and operational resilience. The model's adaptability makes it applicable to fraud detection and predictive maintenance in other industries.