I build open source infrastructure for autonomous AI systems. My work focuses on production-ready patterns that make AI agents reliable, governed, and actually useful in enterprise environments.
Agent OS
A Safety-First Kernel for Autonomous AI Agents. POSIX-inspired primitives with 0% policy violation guarantee. The operating system layer that treats LLMs as raw compute and provides deterministic governance.
Key Features:
- Process Model: POSIX-inspired agent lifecycle management
- Permission System: Fine-grained capability-based access control
- Policy Enforcement: Deterministic rules, not probabilistic guardrails
- Safe Execution: Sandboxed execution with rollback capability
- The Mute Agent: Returns NULL instead of hallucinating
🎯 Results: 0% policy violations vs 26.67% for prompt-based safety
Python
Kernel
AI Safety
Agentic Architecture
A comprehensive guide to modern agentic system design. Documents revolutionary architectural patterns for building production-grade AI agent systems based on real-world experience.
Core Concepts:
- The Inference Trap: Why "thinking" is technical debt
- The Guardrail Router: Intelligent request routing
- Compute-to-Lookup Ratio: 90/10 rule for performance
- Multidimensional Knowledge Graphs: Semantic firewalls
- The Headless Agent: Silent swarms for coordination
- Recursive Ontologies: Self-updating knowledge systems
"If your agent is 'thinking' for every request, you haven't built an agent; you've built a philosophy major. In production, we need engineers, not philosophers."
Architecture
Documentation
Best Practices
Scale by Subtraction
My core methodology for building reliable AI systems. Focus on removing complexity rather than adding features. Via Negativa applied to software architecture and AI safety.
Key Principles:
- Control Planes over Prompts: Deterministic enforcement
- Graphs over Context: Prevent hallucinations structurally
- Silent Swarms over Chat: Structured data communication
- Memory Hygiene: Agents that know how to forget
Methodology
AI Safety
Architecture
Technical Articles
Deep dives into agentic systems, architectural patterns, and the philosophy of building AI that works. Published on Medium, Dev.to, and LinkedIn with practical, battle-tested approaches.
Featured Series:
- The Accumulation Paradox: Why agents degrade over time
- The Mute Agent: Shut up and listen to the graph
- The Agentic Architect: Building AI governance systems
Medium
Dev.to
LinkedIn