Sécuriser les infrastructures critiques. Construire l'IA de demain.
name: Ayi NEDJIMI
role: Senior Cybersecurity Consultant & AI Engineer
certification: OSCP (Offensive Security Certified Professional)
location: France
company: Ayi NEDJIMI Consultants
website: https://ayinedjimi-consultants.frExpert in offensive and defensive cybersecurity with deep specialization in Active Directory security, AI/ML applied to security, and high-performance C++/CUDA development. I build tools that sit at the intersection of cybersecurity, artificial intelligence, and systems programming.
- 20+ years securing critical infrastructures, 50+ audits completed
- OSCP certified — hands-on offensive security expertise
- Specialized in Active Directory, Microsoft 365, Azure, AWS & Kubernetes
- Building AI-powered security tools and fine-tuning domain-specific LLMs
- Publishing 100+ technical articles on cybersecurity and AI
A growing collection of 100+ security and AI tools I develop and maintain. Organized by domain, each tool is built to solve real-world problems I encounter during audits, research, and production deployments.
AI-powered tools that augment security operations — from threat intelligence to automated compliance.
| Tool | Description |
|---|---|
| ThreatIntel-GPT | AI-powered threat intelligence analysis and correlation |
| LogParser-AI | Intelligent log analysis with ML-driven anomaly detection |
| ComplianceBot | Automated compliance checking against ISO 27001, RGPD, NIS2 |
| VulnScanner-LLM | LLM-powered vulnerability scanning and remediation advice |
| PhishingDetector-AI | ML-based phishing email and URL detection engine |
| PolicyGenerator-AI | AI-driven security policy generation from compliance frameworks |
| IncidentSummarizer | Automated incident report summarization for SOC teams |
| CVE-Explorer-AI | AI-assisted CVE exploration, impact analysis and prioritization |
| SOC-Assistant | AI assistant for SOC analysts — triage, enrichment, response |
| SecureCodeReview-AI | AI-powered secure code review for OWASP Top 10 vulnerabilities |
Purpose-built tools for deep auditing and attack detection in Active Directory environments.
| Tool | Description |
|---|---|
| LDAPRecon-AI | AI-powered LDAP enumeration and security audit |
| ACLAudit-AI | Active Directory ACL analysis — dangerous permissions detection |
| KerberosAudit-AI | Kerberos configuration audit and weakness identification |
| GoldenTicket-Detector | Golden Ticket and Silver Ticket attack detection |
| LateralMovement-Detector | Lateral movement pattern detection across the domain |
| RemoteExec-Auditor | Remote execution audit — PsExec, WMI, WinRM, DCOM |
| PrivEscAudit-AD | Privilege escalation path audit in Active Directory |
| DelegationAudit-AD | Kerberos delegation audit — unconstrained, constrained, RBCD |
| DCSyncAudit-AD | DCSync rights audit and replication permission analysis |
| CredentialAudit-AD | Credential hygiene audit — password age, reuse, service accounts |
High-performance GPU-accelerated tools and MLOps infrastructure for production AI pipelines.
| Tool | Description |
|---|---|
| CUDAEmbeddings | GPU-accelerated embedding server with batched inference |
| GPUQuantizer | Model quantization with CUDA — GPTQ, AWQ, GGUF |
| VRAMSwapper | Intelligent VRAM/RAM offloading with multi-stream CUDA transfers |
| ADBloodHound-AI | Active Directory graph analysis with AI-driven path finding |
| YaraGen-AI | AI-powered YARA rule generation from malware samples |
| KQLHunter | KQL query generation for Microsoft Sentinel threat hunting |
| ModelBench | LLM benchmarking suite — latency, throughput, accuracy |
| DatasetForge | Automated dataset creation and augmentation pipeline |
| HashCracker-GPU | GPU-accelerated hash analysis and password audit tool |
| PacketSniffer-AI | ML-driven network traffic analysis and threat detection |
A comprehensive suite of high-performance C++ security tools covering the full audit spectrum:
|
Active Directory Protocols
|
Windows Security
|
Network Protocols
|
All tools are built in modern C++ (C++17/C++20/C++23) with a focus on performance, safety, and minimal dependencies.
KVortex is a production-grade C++23 VRAM-to-RAM KV-Cache offloader for vLLM, enabling larger context windows and higher throughput on GPU-constrained hardware.
- Multi-stream CUDA transfers — overlapped compute and data movement
- Lock-free concurrent queues — zero-contention cache management
- Adaptive eviction policies — LRU, LFU, and frequency-based strategies
- vLLM integration — drop-in plugin for production inference servers
- Benchmarked — measurable throughput gains on real-world LLM workloads
Architecture: CUDA C++23 | Lock-Free Queues | Multi-Stream Transfers | vLLM Plugin
FlashQuant is a production-grade C++17/CUDA implementation of TurboQuant (Google Research, 2025) for extreme KV cache compression in LLM inference.
- 7.5x KV cache compression with less than 2% quality loss
- 6 native CUDA kernels — fused compress, FlashAttention-2, Split-K paged decode
- 12,000+ lines of C++/CUDA/Python with 264 tests passing
- vLLM backend plugin — drop-in attention backend (v0.18-0.22)
Architecture: C++17 | CUDA | PyTorch | pybind11 | FlashAttention-2 | Split-K FlashDecoding
4 Fine-Tuned Models · 50+ Datasets
| Model | Description |
|---|---|
| CyberSec-Assistant-3B | 3B parameter model fine-tuned for cybersecurity Q&A, incident analysis, and threat assessment |
| RGPD-Expert-1.5B | Specialized in GDPR/RGPD compliance — data protection questions and regulatory guidance |
| ISO27001-Expert-1.5B | Fine-tuned for ISO 27001 implementation, audit preparation, and control mapping |
| M365-Expert | Microsoft 365 security configuration, Conditional Access, and administration assistant |
All models are trained on curated, domain-specific datasets built from my consulting experience and technical writing. Available on HuggingFace.
Languages
AI & ML Frameworks
Security & Infrastructure
Cloud & DevOps
- OAuth 2.0 : Sécuriser vos APIs
- Attaques par Relais NTLM — Techniques, Détection et Mitigation
- Sécurité Kubernetes en Production
- Memory Forensics avec Volatility
- Qu'est-ce qu'un Embedding en IA ?
- RAG : Retrieval-Augmented Generation
- Optimiser le Chunking de Documents
- Choisir sa Base Vectorielle — Milvus vs Qdrant vs Weaviate
Browse all 100+ articles at ayinedjimi-consultants.fr/tous-articles.html
| 🌐 Website | ayinedjimi-consultants.fr |
| linkedin.com/in/ayi-nedjimi | |
| 🤗 HuggingFace | huggingface.co/AYI-NEDJIMI |
| 🐦 X / Twitter | @AyiNEDJIMI |
© 2026 Ayi NEDJIMI — All rights reserved
Building security tools by day. Training models by night.