Creative Technologist · AI Product Engineer · Security-minded Builder Arnhem, Netherlands · English C1 · Spanish native · Dutch A2 · Available now — on-site Gelderland, remote EU or relocation
I build products where design, frontend engineering and AI meet — and I verify what I ship with tests, threat models and honest limits.
| Project | What it does | Tests | Stack |
|---|---|---|---|
| osint-evidence-ledger | Tamper-evident evidence ledger for OSINT investigations: SHA-256 chain, NATO Admiral reliability grading, multilingual reporting | 14 | Python · stdlib only |
| eu-sanctions-screen | Explainable name screening against the EU Consolidated Financial Sanctions List with Cyrillic transliteration and scoring transparency | 15 | Python · stdlib only |
| eu-cti-brief | Prioritised vulnerability bulletin from NCSC-NL, CERT-EU, BSI, CISA KEV and EPSS — three priority tiers, watchlist support | 8 | Python · stdlib only |
| citation-audit | Checks whether sources cited in AI-generated answers actually say what the answer claims — no AI inside | 6 | Python · stdlib only |
| chronolocate | Sun position, shadow-based time solving, EXIF consistency, perceptual hashing, Content Credentials (C2PA) detection | 10 | Python · Pillow |
| threat-feed-correlator | Cross-correlate multiple CERT/CISA feeds into one deduplicated, ATT&CK-mapped, EPSS-enriched threat report | 7 | Python · stdlib only |
| ioc-evidence-store | Offline-first IOC database with STIX 2.1 export, SHA-256 integrity chain, CSV/JSON import | 12 | Python · stdlib only |
| web-exposure-scan | Passive e-mail and web exposure check for SMBs — SPF, DKIM, DMARC, TLS, headers — with client-ready reports in NL/EN/ES | 18 | Node · zero deps |
| Project | What it does | Stack | Live |
|---|---|---|---|
| mcp-osint-server | MCP server exposing OSINT tools (sanctions screening, CTI briefs, IOC search, chronolocation) to AI agents | TypeScript · MCP SDK | - |
| civil-sentry | Evidence-driven cyber situational awareness: authorization gate, SHA-256 evidence chain, AI findings grounded in proof | TypeScript · zero runtime deps · Apache-2.0 | site |
| Buyer Arena | Evidence-first evaluation of web products and AI-agent changes: seeded synthetic buyers in a real browser, baseline vs candidate, release gates | TypeScript · Node 22 · Playwright | - |
| npm-supply-chain-auditor | Read-only scanner for compromised npm packages, droppers and persistence hooks | PowerShell · offline IOC dataset | - |
| ARGUS | Evidence and opportunity control plane: typed, hashed evidence from collectors, correlated into findings and retests. 195 tests | TypeScript · pnpm monorepo | - |
| Project | What it does | Stack | Live |
|---|---|---|---|
| NEXUS Visual Engine | 30 real-time audio-reactive visual engines for DJ sets, museums and installations | WebGL · Three.js · GLSL · Web Audio | demo |
| IAQUARIUS Gateway | Local multi-provider AI gateway: one OpenAI-compatible endpoint with automatic failover, chat UI and observability | LiteLLM · Open WebUI · Docker | - |
| Gelderland Crane Simulator | Deterministic overhead-crane training simulator with 35 automated physics checks | Godot · GDScript | - |
| loopsmith | Exact-duration assembly of AI-generated video clips with seam analysis | Bash · FFmpeg/ffprobe | - |
| civic-relay | Offline-first crisis messaging prototype: multi-transport routing, store-and-forward, deduplication | TypeScript · React · pnpm monorepo | demo |
I use AI agents (Claude, Codex, Gemini) as implementation partners, not as authors of record: I set scope and architecture, agents implement, and nothing is claimed as done unless a test or command proves it. Examples of that workflow are public — the crane simulator's .claude/ agent roles and rules, and civic-relay's RELEASE_TRUTH_V0_1.md verification record.
CTI & Threat Intelligence · SOC Engineering · AI Product Engineering · Creative Technology · Design Engineering · Security Engineering (defensive) · Applied AI / Forward Deployed