Personal projects: AI-based detection systems (compromised accounts via behavioural analysis; static malware analysis)
Don't disclose
Created AI models for security-focused classification and anomaly detection using supervised and unsupervised approaches. Prepared and trained models on labeled/tabular features to produce probabilistic outputs for downstream analysis. Implemented end-user tooling for interaction and visualization of predictions. • LSTM autoencoder anomaly detection using keystroke dynamics features (Hold Time, Down-Down Time, Up-Down Time) • Random Forest malware classifier trained on pre-extracted PE header metadata from labeled CSV • Calibrated thresholds using ROC-based methods to control detection rates • Mapped model predictions to MITRE ATT&CK tactics and techniques for interpretability
2023