# Clint Josy

AI Security Researcher & Leader. AI-driven security · offensive security · exposure management.

Pushing AI further. Engineering security differently.

## Profile
- name: Clint Josy
- role: AI security researcher, leader
- location: Bengaluru, India
- focus: offensive security, exposure management, AI-driven defense

## About
- background: authorized offensive security, then remediation with the owners
- current focus: AI for security teams, and the security of AI systems
- approach: identify what matters, explain it plainly, get it fixed
- scope: applications, infrastructure, cloud, APIs, models

## Focus
- offensive security: Red team operations, Web, API and cloud assessments, Active Directory security, Social engineering awareness, Vulnerability research
- ai security: AI and ML penetration testing, LLM security, AI risk management (NIST AI RMF), AI-assisted detection
- exposure management: Attack-surface mapping, Threat modelling, Cyber threat intelligence, Risk prioritization, Digital forensics
- platforms: Microsoft Azure and Entra ID, Proxmox and self-hosted LLMs, Linux, Docker, NVIDIA, Python, Bash, TypeScript

## Selected work
- [One Password, No Device, Full Tenant](https://www.cyderes.com/howler-cell/azure-ad-conditional-access-device-identity-abuse) (research): Entra ID Conditional Access; device identity trust assumptions; hardening guidance
- [A forgotten certificate authority](https://www.linkedin.com/feed/update/urn:li:activity:7476270373144182785/) (research): Active Directory; ungoverned certificate authority; Zero Trust blind spot; defender priorities
- [CVE-2019-8920: XAMPP 1.7.0 reflected XSS](https://medium.com/@0xm4rv/exploit-title-xampp-1-7-0-reflected-cross-site-scripting-782ddb02d312) (disclosure): Reflected XSS; XAMPP 1.7.0; CVE-2019-8920; CVSS 6.1 Medium; NVD listed
  - [nvd entry](https://nvd.nist.gov/vuln/detail/cve-2019-8920)
- [Local LLM with GPU acceleration on Proxmox](https://clintjosy.pages.dev/blog/2026-05-01-your-ai-your-rules-running-a-local-llm-with-gpu-acceleration-on-proxmox) (build): Proxmox LXC; NVIDIA passthrough; MoE model; 3 tok/s to 21 tok/s; full guide
- [OpenMythos teardown](https://clintjosy.pages.dev/blog/2026-04-23-openmythos-teardown-dissecting-the-open-source-reconstruction-of-claude-mythos) (write-up): Recurrent-depth transformer; architecture + training pipeline; security implications of local reasoning models

## Experience
- Jul 2023 to Aug 2026: Offensive Security Lead, Security Strategist, Cyderes, Bengaluru, India. Lead offensive security practice; exposure management engagements; Howler Cell research (identity, cloud access)
- Feb 2021 to Jul 2023: Senior Security Engineer, Axiata, Kuala Lumpur, Malaysia (remote). Application security; web architecture review; data privacy; telecom group, multi-country
- Sep 2018 to Jan 2021: Security Analyst, Safe Security, New Delhi, India. Intern then analyst; network and application testing; vulnerability research; cyber risk quantification

## Certifications
- 2019: Offensive Security Certified Professional (OSCP) (OffSec)
- 2020: Certified Red Team Operator (CRTO) (Zero-Point Security)
- 2025: Computer Hacking Forensic Investigator (CHFI) (EC-Council)
- 2025: Certified AI/ML Pentester (C-AI/MLPen) (The SecOps Group)
- 2025: Security Operations Analyst Associate (SC-200) (Microsoft)
- 2025: Foundations of AI Security (AttackIQ)
- 2023: Certified AppSec Practitioner (CAP) (The SecOps Group)
- 2023: Certified Network Security Practitioner (CNSP) (The SecOps Group)
- 2022: Azure Fundamentals (AZ-900) (Microsoft)
- 2020: Social Engineering Expert (RedTeam Security Training)

## Writing
- 2026-05-01: [Your AI, Your Rules: Running a Local LLM with GPU Acceleration on Proxmox](https://clintjosy.pages.dev/blog/2026-05-01-your-ai-your-rules-running-a-local-llm-with-gpu-acceleration-on-proxmox)
- 2026-04-23: [OpenMythos Teardown: Dissecting the Open-Source Reconstruction of Claude Mythos](https://clintjosy.pages.dev/blog/2026-04-23-openmythos-teardown-dissecting-the-open-source-reconstruction-of-claude-mythos)

## Contact
- topics: AI security, research, collaboration
- channel: LinkedIn
- also: dev.to, GitHub
- [LinkedIn](https://www.linkedin.com/in/clintjosy/)
- [dev.to](https://dev.to/clintjosy)
- [GitHub](https://github.com/m4rvxpn)