Mahdi Aouiti

Worked at the intersection of DevSecOps and Machine Learning to improve application security, observability, and performance within CI/CD pipelines at Teligencia. Key achievements: - Built and secured end-to-end CI/CD pipelines (Jenkins, Docker) for React/Node.js apps, reducing deployment failures by 40% via automated tests and security scanning. - Designed and deployed a real-time monitoring infrastructure (Prometheus, Grafana, Alertmanager) with critical alerts (CPU >85%, API error rate >5%), cutting incident resolution time by 30%. - Developed a machine learning model (Random Forest) to detect anomalies in API latency and system metrics, improving detection of early performance issues. -Trained the model on historical Prometheus metrics and integrated it with alert rules for predictive monitoring. - Applied ML to optimize alert thresholds and reduce false positives in monitoring systems. - Integrated Shift-Left security tools (SonarQube, OWASP Dependency Check, Snyk) into the pipeline, identifying and fixing 90% of vulnerabilities pre-production. - Created CLI tools in TypeScript to automate Docker and dependency audits, reducing manual workload by 50%.

I designed and developed a smart, AI-based Network Intrusion Detection System (NIDS) as part of LAB619’s cybersecurity R&D. This solution uses machine learning (Random Forest, SVM) to analyze live network traffic and automatically detect malicious behaviors, especially in IoT and industrial systems. 🔧 Key features: • Dataset training with NSL-KDD and CICIDS2017 • Feature extraction, normalization, and model training • Real-time detection using lightweight agents on edge devices (Raspberry Pi) • API alerts with FastAPI backend and Grafana dashboard for visualization 🎯 Impact: This project enhanced LAB619’s IoT Acceleration service by embedding proactive cybersecurity into custom devices and industrial machines, aligned with Industry 4.0 best practices.

Designed and deployed a network infrastructure featuring two cascading firewalls (PfSense and IPFire) across three security zones: WAN, LAN, and DMZ. - Ensured secure connectivity and segmentation. - Implemented a SIEM and SOAR system with orchestration capabilities using tools such as ELK Stack, Wazuh, Security Onion, Suricata, Cortex, and The Hive. - Integrated Shuffle for automation and alert response workflows. - Deployed honeypots to detect malicious activities. - Conducted a full vulnerability audit using tools like Nmap and Nessus, followed by a comprehensive security report. 🛠 Technologies used: VMware Workstation, VirtualBox, PfSense, IPFire, ELK Stack, Wazuh, Suricata, The Hive, Cortex, Shuffle, Security Onion, Ubuntu, CentOS, Windows, Honeypot, Nmap, Nessus.

- Designed and developed a secure RESTful API to automate secrets lifecycle management using HashiCorp Vault, ensuring encrypted storage and granular access control for sensitive credentials across multi-cloud and hybrid environments. - Integrated real-time security alerts with Slack via webhook APIs and monitoring tools (Prometheus/Grafana), enabling instant team notifications for critical security events or compliance violations. - Reduced container vulnerabilities by 60% by implementing automated scans (Trivy/Anchore) in CI/CD pipelines, driving continuous remediation of exploitable risks. - Embedded security into DevOps workflows, collaborating cross-functionally to enforce security best practices (SBOM, IaC scanning) without compromising deployment velocity. - Documented security protocols for incident response and knowledge sharing, streamlining team onboarding and reducing mean-time-to-resolution (MTTR) by 30%.

- Conducted academic research on the robustness and security of machine learning models, with a focus on adversarial machine learning and bias mitigation. - Explored the impact of adversarial examples on classification accuracy and investigated regularization, data augmentation, and ensemble methods as defense mechanisms. - Designed and implemented Python-based tools to monitor key ML performance metrics such as precision, latency, and data drift under attack scenarios. - Gained hands-on experience with academic experimentation workflows, including dataset preprocessing, model evaluation, and reproducibility of results. - Contributed to a research-driven approach to secure AI development, laying the groundwork for further studies in applied machine learning and cybersecurity.