Skip to content
View Mahad7836's full-sized avatar

Block or report Mahad7836

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Mahad7836/README.md

Mahad Aqeel

AI Security · LLM Evaluation · Security Automation · Offensive Security

Final-year BS Cybersecurity student at GIKI (Class of 2027) focused on evaluating and securing LLM and agentic systems, building practical security automation, and applying offensive-security thinking to emerging AI systems.

LinkedIn · Email


Focus

  • LLM & Agent Security — prompt injection, adversarial evaluation, tool-use security, attack/defense evaluation
  • Security Automation — AI-assisted triage, workflow automation, policy enforcement, alerting
  • Applied AI for Security — model evaluation, classification, explainability, multilingual LLM analysis
  • Offensive Security — VAPT, web/network security, CTFs, vulnerability analysis

Selected Work

LLM Security · Cross-Lingual Evaluation · Reproducible Research

Reproducible evaluation of cross-lingual prompt-injection and protected-context leakage in aligned large language models.


Python · NVIDIA garak · Mistral-7B · Adversarial Evaluation

Evaluated an instruction-tuned LLM across 64 adversarial attempts, establishing a 35.94% attack-success baseline and extending the evaluation to English–Urdu prompt-injection behaviour.


n8n · Security Automation · Webhooks · Slack

Policy-governed AI-security workflow for structured alert ingestion, payload validation, risk/severity routing, logging, and automated incident-response notifications.


Python · XGBoost · TF-IDF · SHAP

Machine-learning pipeline for detecting sensitive information using character-level features, metadata, and explainable classification, achieving 94% classification accuracy in project evaluation.


JavaScript · Browser Privacy · Fingerprinting

Browser-extension project exploring resistance to browser-fingerprinting and behavioural profiling through controlled modification of browser-identifying attributes.


Web · Cryptography · Reverse Engineering · Forensics · OSINT

Curated technical writeups from an ongoing offensive-security journey spanning 80+ CTF challenges across multiple security domains.


Current Research & Engineering

Adaptive Attack–Defense Co-Evaluation for Enterprise LLM Chatbots

Final-year project exploring adaptive attacker and defender agents, shared-state coordination, prompt injection, tool-use security, and measurable attack/defense evaluation for enterprise LLM systems.

Development repository is currently private while the project is in active development.


Experience

Information Security Analyst Intern — RICI Company Ltd.

Dammam, Saudi Arabia · Jun–Aug 2025

Conducted enterprise VAPT across 150+ hosts, identified and prioritized high-risk exposures, contributed to ISO 27001:2022 compliance work, and translated technical findings into remediation guidance for technical and management stakeholders.

AI Campus Ambassador — Confinality

Remote · Dec 2025–Present

Support AI-focused campus engagement and serve as a liaison between Confinality and the GIKI community around LLM systems, AI assistants, and emerging AI technologies.


Technical Stack

Languages: Python · C++ · JavaScript · SQL · Bash

AI / ML: NVIDIA garak · scikit-learn · XGBoost · Hugging Face · SHAP · LLM Evaluation

Security: Nmap · Nessus · OWASP ZAP · Nikto · Burp Suite · VAPT · OWASP Top 10

Automation & Engineering: n8n · Webhooks · APIs · Git/GitHub · Linux · Docker

Governance: ISO 27001:2022 · Security Policy Development · Risk Assessment


Beyond Code

  • Top 8% globally on TryHackMe
  • Founder & President of The Ansars, leading multi-team humanitarian and community initiatives
  • Technical writer covering cybersecurity, AI security, research, and practical learning
  • Research interests: LLM security, agent security, adversarial evaluation, multilingual AI safety

Contact

I'm interested in opportunities and collaborations involving AI security, LLM evaluation, security automation, offensive security, applied AI engineering, and technical AI/security products.

Email: mahad.aqeel7@gmail.com
LinkedIn: linkedin.com/in/mahad-aqeel

Pinned Loading

  1. AdvSoc AdvSoc Public

    Policy-governed AI security automation workflow for threat triage, severity routing, structured logging, and incident-response alerting using n8n.

    Python

  2. AlignSec AlignSec Public

    Reproducible evaluation of cross-lingual prompt-injection protected-context leakage in aligned large language models.

    Jupyter Notebook

  3. DLP-Project DLP-Project Public

    Machine-learning pipeline for PII detection and data-loss prevention using XGBoost, TF-IDF character features, metadata, and SHAP.

    Python 1

  4. garak-multilingual-promptinj-eval garak-multilingual-promptinj-eval Public

    Reproducible English–Urdu prompt-injection evaluation with NVIDIA garak, automated detectors, and paired response review.

    Python

  5. Secure-AI-Prompt-Sandbox Secure-AI-Prompt-Sandbox Public

    Forked from Deez-Automations/Secure-AI-Prompt-Sandbox

    A secure LLM gateway that detects prompt injection, enforces security policies, and integrates usable security features such as real-time risk scoring and transparent prompt validation.

    JavaScript