AI Security Engineer | Agentic AI Security | LLM Red Teaming | MCP Security | AI Governance
Houston, TX | Ibrahimsaleem244@gmail.com
I am an AI Security & Governance Engineer focused on securing enterprise-scale LLM, GenAI, MCP, and agentic AI systems.
My work focuses on practical AI security risks such as prompt injection, jailbreaks, tool poisoning, sensitive data leakage, excessive agency, insecure MCP servers, unsafe RAG, and missing runtime controls.
I work across AI threat modeling, adversarial testing, secure AI SDLC, AI governance, red teaming, and runtime security controls for enterprise AI systems.
- LLM and GenAI security
- Agentic AI and MCP server security
- AI red teaming and adversarial testing
- Secure RAG and data leakage prevention
- AI governance and product risk reviews
- Secure AI SDLC and runtime monitoring
- OWASP Top 10 for LLMs, NIST AI RMF, ISO/IEC 42001
Master of Science in Cybersecurity
Houston, TX | Expected May 2026
GPA: 3.98/4.0 | Awarded $16K Scholarship
Relevant Coursework: Network Security, Secure Enterprise Computing, Cryptography, Data Analysis for Cybersecurity, Cybersecurity Risk Management, Secure Software Design
Bachelor of Technology in Computer Science Engineering
July 2023
AT&T — Plano, TX / Hybrid
Jan 2026 – Present
- Securing enterprise GenAI and agentic AI systems through AI governance workflows, threat modeling, and runtime control requirements.
- Building AI security review frameworks aligned with OWASP LLM Top 10, NIST AI RMF, ISO/IEC 42001, and secure AI SDLC practices.
- Performing adversarial testing and red-team assessments on enterprise LLM systems before deployment.
- Developing privacy-preserving tokenization and masking workflows for sensitive enterprise networking data.
NOV Inc. / National Oilwell Varco — Houston, TX
June 2025 – Dec 2025
- Built secure GenAI automation systems using LLMs, OCR, RAG, agent orchestration, and automated validation.
- Secured AI workflows against prompt injection, jailbreaks, sensitive data leakage, excessive agency, and tool misuse.
- Developed and secured MCP-based systems for enterprise automation use cases.
- Contributed to a published SPE research paper on self-improving GenAI agents for automated report parsing.
University of Houston — Houston, TX
Sep 2024 – May 2025
- Led research on LIMA, an LLM-powered penetration-testing framework using MCP servers for initial machine access.
- Built PentestThinkingMCP, an MCP-based AI reasoning server for autonomous penetration-testing workflows.
- Researched AI-assisted offensive security, attack-path planning, and secure MCP tool execution.
Nagarro Software Pvt. Ltd.
Mar 2023 – Feb 2024
- Built secure backend systems using C#, .NET Core, SQL Server, REST APIs, JWT, RBAC, and secure coding practices.
- Worked on API security, input validation, SQL injection prevention, and enterprise application development.
| Project | Focus |
|---|---|
| PentestThinkingMCP | MCP-based AI penetration-testing reasoning server |
| LocalRAGAgent | Offline privacy-preserving RAG pipeline |
| Secure Offline AI Assistant | On-prem LLM assistant for network operations |
| ML DDoS Detection | ML-based network monitoring and DDoS detection |
| Cybersecurity Awareness Platform | Hands-on cybersecurity learning platform |
-
LIMA: Leveraging Large Language Models and MCP Servers for Initial Machine Access
IEEE FMLDS 2025
Google Scholar Search | ResearchGate -
Self-Improving Generative AI Agents for Automated Daily Mud Report Parsing
IADC/SPE International Drilling Conference and Exhibition 2026
Google Scholar Search | OnePetro Search
AI Security: LLM security, prompt injection defense, jailbreak testing, secure RAG, AI red teaming, MCP security, agentic AI threat modeling, runtime controls
Governance & Frameworks: OWASP Top 10 for LLMs, OWASP Agentic AI Threat Modeling, NIST AI RMF, ISO/IEC 42001, Secure AI SDLC
Security Tools: Burp Suite, Metasploit, Nmap, Wireshark, Splunk, Cortex XSOAR, Semgrep, GitLeaks, Trivy
Programming: Python, C#, C++, SQL, Bash, PowerShell, JavaScript/TypeScript
AI / ML / LLM Tools: LangChain, LangGraph, Hugging Face, Llama/Ollama, Azure OpenAI, FastAPI, Flask, Streamlit
Cloud & DevSecOps: Azure, AWS, Docker, Kubernetes, GitLab CI/CD, Databricks, Azure Key Vault, Managed Identities, Defender for Cloud
AI systems introduce new trust boundaries.
My approach:
- Treat prompts as untrusted input.
- Treat retrieved context as potentially poisoned.
- Give agents least privilege.
- Monitor tool use and runtime behavior.
- Build AI systems that are secure, auditable, and controllable.
- Portfolio: ibrahimsaleem.com
- Project Portfolio: ibrahimsaleem-portfolio.web.app
- LinkedIn: linkedin.com/in/ibrahimsaleem91
- GitHub: github.com/ibrahimsaleem
- Google Scholar: Mohammad Ibrahim Saleem
- Email: Ibrahimsaleem244@gmail.com
Building secure AI systems means controlling what the model can see, what the agent can do, what the tools can access, and what the enterprise can prove afterward.