I build production-oriented tools at the intersection of AI and security.
Focus areas: prompt injection & jailbreak defense, secrets detection with context-aware false-positive reduction, passive attack-surface mapping, and automated CVE/SIEM triage.
I also write first-principles technical deep-dives on LLM infrastructure (KV cache, Ring Attention, quantization, speculative decoding) and AI security (RAG prompt injection, vector search internals).
| Project | Description |
|---|---|
| llmguard-cli | Real-time prompt injection + jailbreak detector. Multi-layer: 20+ heuristic signatures + LLM meta-reasoning. Includes HTTP API for pipeline integration. |
| gitleaks-ai | AI-enhanced secrets scanner. Shannon entropy analysis + LLM false-positive elimination (~73% fewer FPs vs regex-only). Drop-in upgrade path for existing scanners. |
| shadowmap | Passive attack surface mapper. Certificate transparency + DNS enumeration + AI risk scoring. Zero active scanning. |
| cvewatch | CVE monitoring daemon. NVD API 2.0 + AI relevance scoring against your stack + Slack/Discord alerts. |
| yaraforge | AI-powered YARA rule generator and optimizer. Generate, validate, and deploy to Elastic / Splunk / standalone YARA from one CLI. |
| ai-siem-triage | SOC alert triage assistant. Classifies SIEM events, prioritizes by severity, and recommends response actions. |
- Rust CLI tools:
rust-grep-tool,rust-sys-monitor,rust-config-transformer— focused systems programming practice. - Mechanistic interpretability: nano-mind-reader — visualizing internal patterns of a tiny GPT.
- DeFi / crypto research: stablecoin de-peg early warning and macro-crypto correlation experiments.
Deep technical posts on LinkedIn covering LLM systems and AI security from first principles:
- Ring Attention & long-context architectures
- KV Cache, quantization, speculative decoding
- Prompt injection in RAG agents
- Vector search internals
- Distributed systems fundamentals (Kafka, etc.)
The best security tool is the one that fits your workflow.