Selected projects spanning AI systems, embedded development, full-stack applications, and open-source products.
Featured
Codex Neuralis — Company & Portfolio
My company. We build AI systems, AI agents, and software using AI-native infrastructure. We design and deploy multi-agent orchestrations, autonomous agent systems, custom MCP plugins and proprietary skill sets, agentic CLI and TUI tools, embedded systems with AI copilots, automated content pipelines, and open-source products that serve the developer community worldwide. Every project carries the mark: Built by Codex Neuralis.
Our portfolio spans enterprise consulting, production AI deployments, audio podcast production, editorial content generation, and startup acceleration across the European ecosystem. We operate on a single standard: uncompromising quality, production-grade from day one, with observability, security, and scalability built into every deliverable.
An autonomous AI agent operating 24/7, built by Codex Neuralis. Mihail Gepetescu runs from Hermes Agent, Claude Code, Codex, and Opencode.ai Deepseek TUI, executing proprietary skill sets that control his entire web presence autonomously via cronjobs — researching, writing, publishing, and managing content without human intervention. Fully bilingual (Romanian/English), the agent powers mihailgepetescu.pro with blog, podcast, photo gallery, and poetry, all published on schedule by the agent itself.
Key capabilities: autonomous news writing with multi-source research, podcast script generation and production, software development in TypeScript, Python, Rust, and Go, GitHub repository management, and multi-domain publishing. The agent operates with zero human intervention — a demonstration of Codex Neuralis' agentic AI infrastructure in production, deployable across multiple agent runtimes.
Real-time 802.11 packet capture and threat analysis agent running on an ESP32-S3 Zero (Waveshare, 4MB Flash, 2MB PSRAM). Firmware written in C using ESP-IDF captures Wi-Fi probe requests, beacon frames, and deauthentication packets. A Python agent backend processes the data, runs an LLM copilot for threat analysis, and serves a live dashboard with real-time statistics, device tracking, and AP mapping.
Key features: 802.11 packet sniffing in monitor mode, deauth detection with alerting, periodic AP reporting with BSSID/SSID/channel/RSSI, SQLite persistence, REST API, and an AI copilot that answers questions about the current network state.
ESP-IDFCPythonAI AgentSQLiteREST API
AI Agent Orchestration
Multi-agent systems that automate complex research, content generation, and data processing pipelines. Custom architectures using Python, LangChain, and direct LLM orchestration. Agents collaborate on tasks such as deep web research with source verification, article drafting with fact-checking, audio podcast production with multi-voice synthesis, and magazine layout generation with visual effects.
Each pipeline is designed for reliability - agents verify each other's work, retry on failure, and produce consistent output suitable for production delivery. n8n workflows connect the agent pipelines to external services, APIs, and deployment targets.
PythonLangChainLLMsn8nREST APIsElevenLabs
TUI & Shell Agent Tools
Terminal-based AI agents and CLI tools designed for developers who work from the command line. These are agentic interfaces that combine shell automation with LLM-powered reasoning - enabling complex multi-step operations triggered by natural language commands.
Built for power users who need AI assistance without leaving the terminal. TUI agents provide real-time feedback, streaming responses, and direct system integration through secure command execution patterns.
PythonTUICLILLMsShell
n8n Automation Pipelines
Complex automation workflows connecting AI agents, web services, databases, and APIs. n8n serves as the orchestration layer for content publishing pipelines, data extraction and enrichment workflows, multi-platform social media automation, and client onboarding sequences.
Pipelines are designed for reliability with error handling, retry logic, and notification systems. They bridge the gap between AI agent output and production delivery to WordPress, custom APIs, email, and storage systems.