Kader Mohideen
Senior Presales Consultant
Starlink MDMC
I’m a Senior Presales Consultant at Starlink DMCC and an active open-source researcher working at the intersection of AI and security. I advise banks, telecoms, and government entities in Saudi Arabia on designing GenAI, Zero Trust, and SOC programmes aligned with NCA, SAMA, and SDAIA requirements. I have led technical evaluations of more than 60 enterprise security products across the GCC.
My public projects include AddLM, a multiplication-free transformer experiment; Employee Recall, which explores persona continuity using LoRA and RAG; DoodleCode Studio; the 31-module Data Science Roadmap; AI Security Notes; and the Attack-Centric Framework, a defensive model that centres security strategy on attacker behaviour.
Explore my work at kader-xai.github.io.
Speaker sessions
Briefings and Arsenals | Confirmed Speakers | No Session Assigned
Placeholder session for confirmed speakers without an assigned session, allowing them to be pushed to Swapcard and Visit.
- 07:30
- Tue
- 01 Dec
Stage:
Briefing Stage 2
Sessions Type:
Keynote
OWASP-AI-Range: A Vulnerable-by-Design LLM Stack for Hands-On AI Red Teaming and Guardrail Benchmarking
Every CISO in the region is being asked the same question: "are we ready to red-team our AI?" Yet unlike web application security � where defenders cut their teeth on DVWA, Juice Shop and HackTheBox � there is no standard, open environment to learn or measure offensive AI skills. Practitioners read the OWASP Top 10 for LLM Applications and the MITRE ATLAS matrix, then have nowhere to actually practice the attacks against a realistic target.OWASP-AI-Range closes that gap. It is an open-source, vulnerable-by-design LLM application stack that implements every category of the OWASP LLM Top 10 (2025) as a hands-on lab, plus extended scenarios for agentic systems, retrieval-augmented generation (RAG) pipelines and Model Context Protocol (MCP) servers. Each lab ships with a deliberately broken application, a capture-the-flag objective, a graded scoring harness, a "fix" branch demonstrating the secure pattern, and instructor notes mapped to MITRE ATLAS techniques.The range runs locally via Docker Compose with pluggable model backends (OpenAI, Anthropic, local Llama and Mistral) so security teams can train without cloud spend or data exposure. A built-in CI harness lets organisations measure red-team progress over time and benchmark commercial guardrails (Llama Guard, Prompt Shields, Lakera, NeMo Guardrails) against a reproducible attack suite.This session live-deploys the range, walks attackers through three flagship exploits end-to-end � indirect prompt injection, RAG cache poisoning, and MCP tool poisoning � and shows the scoring leaderboard comparing five commercial guardrails. Attendees leave with the full toolkit and a turnkey AI red-team blueprint.
- 14:50
- Wed
- 02 Dec
Stage:
Briefing Stage 2
Sessions Type:
Keynote