Principal AI Systems Architect | Agentic AI, Voice AI & Intelligence Platforms | 25+ years shipping production systems
Rome, Italy • LinkedIn • X
I design and ship production-grade AI systems: real-time voice agents, agentic workflows, intelligence platforms, semantic retrieval pipelines, and the infrastructure behind them.
GPT Live-style voice AI running directly in your terminal, with a compact Python core of roughly 1,500 lines.
A headless, real-time voice assistant powered by ElevenLabs, Gemini, Mem0, and MCP tools. It combines natural voice conversations, persistent memory, extensible tools, and local wake-word detection without a bloated UI or a complicated local-model stack.
Python • ElevenLabs • Gemini • Mem0 • MCP • Voice AI
AI-powered OSINT platform for intelligence analysts, historical researchers, and investigative journalists.
Built with FastAPI, Next.js, Weaviate, DSPy, GLiNER entity extraction, and Mapbox 3D geospatial visualization.
Origin: built to investigate the Cold War fate of my aunt Ingeborg Loucek, a U.S. Army Counter Intelligence Corps agent who disappeared in Vienna in 1950.
| Repository | Description |
|---|---|
| intellyweave | Core intelligence platform |
| intellyweave-cli | Operations CLI and AI shell |
| mcp-weaviate | MCP server for Weaviate integration |
Design voice agents visually. Generate production-ready code automatically.
GraphForge is a self-hosted, design-first development platform for building voice-enabled AI agents. Teams can draw conversation flows, configure providers visually, export portable workflow specifications, and generate LangGraph-based Python or TypeScript without rewriting infrastructure.
- Drag-and-drop editor with 12 node types for triggers, processing, responses, and control flow
- Visual configuration for OpenAI, Deepgram, ElevenLabs, and Silero
- Production-oriented code generation with customizable Jinja2 templates
- YAML import/export for version control and team collaboration
- Dagre and ELK auto-layout for complex conversation graphs
- Pre-built workflow templates and a LiveKit voice runtime with human-in-the-loop interrupts
Next.js • React • TypeScript • Python • LangGraph • LiveKit • Deepgram • ElevenLabs
Repository • Documentation • Roadmap
- Agentic AI: multi-agent architectures, RAG systems, semantic retrieval, LLM orchestration, and MCP
- Voice AI: real-time conversational systems, WebRTC, speech recognition, synthesis, and telephony
- Intelligence Platforms: OSINT, knowledge graphs, entity extraction, and investigative workflows
- Vector Databases: Weaviate, Pinecone, vector-first data models, and hybrid search
- GIS & Geospatial: 20+ years with ArcGIS, QGIS, Mapbox GL, and geospatial intelligence
- Full Stack: Python, FastAPI, Next.js, React, TypeScript, and Node.js
| Period | Role | Organization |
|---|---|---|
| 2022–Present | Principal AI Systems Architect | Independent Consulting |
| 2018–2022 | Senior System Architect | DXC Technology |
| 2012–2018 | R&D Team Lead | DYMATRIX Germany |
| 2010–2012 | Senior Developer | Hewlett-Packard |
| 2005–2009 | Team Leader | Poste Italiane |
Building AI as infrastructure, not experiments.