Tools I build.

AI-powered network infrastructure management — SSH, NETCONF, NetBox, EVE-NG. Asimov Firewall (dual-LLM safety). Self-hosted, self-learning.

4-layer AI safety system with TCP/IP-style layering. 16 rules across Base Laws, Security, Scope, and Behavioral layers. Deterministic denylist + independent AI judge.

Multi-vendor network automation. Parallel SSH, NETCONF, eAPI, telnet. Commit-confirmed with auto-rollback. Audit trails.

Multi-agent coordination framework: orchestrates Claude, Codex and other agent CLIs against real infrastructure.

Fork of NVIDIA's OpenShell — the safe, private runtime for autonomous AI agents — where h-flock autonomously found and fixed real bugs. Proof it can fix, not just orchestrate.

General-purpose agent dev container: CLI tooling on Ubuntu 24.04, unprivileged, built for amd64/arm64, published to GHCR.

A real-world implementation case for the Asimov Firewall, built on a durable, auditable agent execution substrate — Apache Camel + Temporal + custom gRPC.

Streaming gRPC bridge with OpenTelemetry instrumentation (ramp-on/ramp-off pattern).

Reports and results from LLM testing for infrastructure management.

Network traffic analyser with custom Redis orchestration, NeMo agent training toolkit + Morpheus integration.

Community RFC proposals, including draft-baysal-asimov-safety-architecture.

Published security audits and hardening reports for AI infrastructure.

Ask me about this one.