madKoding
Software Engineer — AI Systems, Systems Programming & Full Stack
Chile
About
Software engineer focused on AI systems, systems-level programming in Rust, and full-stack development. I design and build end-to-end solutions — from custom Linux distributions and embedded firmware to LLM orchestration engines and real-time web applications.
Currently building the NEURO ecosystem: an interconnected platform where madOS provides the AI-integrated OS layer, Fractal-Mind serves as the cognitive memory engine, Neuro Agent orchestrates multi-model AI workflows, and neuro-ollama manages GPU-accelerated local inference.
My work spans the full depth of the stack:
- AI & Knowledge Systems — RAG pipelines, LLM orchestration with multi-provider abstraction, vector embeddings, fractal graph-based memory architectures
- Systems Programming — Custom OS builds with adaptive hardware detection, kernel tuning, ZRAM compression, Wayland compositing
- Full Stack — React/Next.js frontends, Node.js/Express backends, REST APIs with JWT auth, real-time systems with WebSocket
- Infrastructure & DevOps — Docker multi-service architectures, GitHub Actions CI/CD, GPU-accelerated containerized inference
- Embedded & IoT — ESP32 firmware, VR tracking systems (SlimeVR), nRF52 BLE bootloaders, Android native (Kotlin)
Featured Projects
madOS
AI-orchestrated Arch Linux distribution for resource-constrained systems (1.9GB+ RAM). Features adaptive GPU rendering (auto-switches between hardware acceleration and software fallback), ZRAM with zstd compression, audio quality auto-detection up to 192kHz/32-bit, USB persistence with dynamic space allocation, and Sway Wayland compositor at ~67MB RAM footprint. Automated ISO builds via GitHub Actions.
|
Fractal-Mind
AI cognitive engine with evolutionary memory built on RAPTOR-based fractal graph structures and HNSW vector indexing over SurrealDB. Dual-phase architecture: Vigilia (wakefulness) for real-time response via graph navigation, REM (sleep) for asynchronous learning and memory consolidation. Multi-user namespace isolation, PDF/OCR ingestion, multi-embedding support (Nomic, BGE, CLIP). Full roadmap delivered — 12/12 features complete.
|
Neuro Agent
AI programming assistant in Rust with dual-model orchestration — routes tasks between fast local models (Ollama) and heavy compute providers (OpenAI, Anthropic, Groq) through a unified provider abstraction. Features AST-based code review with cyclomatic complexity detection, LRU context caching (10x response improvement), smart error recovery with retry/rollback, and a performance benchmarking framework with regression detection. 219+ tests, 100% pass rate.
|
Git Reports
Automated Git analytics engine in Rust that generates work summaries from repository activity. Analyzes commits, contributors, and activity patterns across time periods with multi-format export. Integrates LLM-powered report generation for natural language summaries. CI/CD with GitHub Actions and 100% passing test coverage.
|
MMORPGVT
Real-time interactive MMORPG engine for Twitch VTuber streams. Chat-driven mechanics: battles, exploration, virtual economy, and collaborative quests — all triggered from viewer commands. Built with Next.js and Firebase for real-time state sync. Full test suite with Jest, ESLint, Prettier, and automated formatting.
|
Docker SlimeVR
Production-ready Docker environment for SlimeVR full-body tracking server with dynamic Web GUI. Auto-downloads server JAR and GUI assets from GitHub releases, serves via Nginx with LAN IP auto-redirect. Exposes tracker (UDP 6969), WebSocket bridge, OSC router, VMC, and VRC ports. Persistent volume management for config and GUI assets.
|
See all repositories at github.com/madkoding?tab=repositories
Tech Stack
Activity
Última actualización: 2026-02-26 15:51:07 -03