Full-Stack Developer and Senior Software Engineer based in Chile, specializing in AI systems, systems programming, and distributed infrastructure. I architect production-grade solutions spanning the entire stack — from kernel-level optimizations and embedded firmware to LLM orchestration engines and real-time web platforms. Beyond engineering, I am also a VTuber and content creator focused on VR technology, full-body tracking, and software development on Twitch and TikTok.
Currently leading the NEURO ecosystem: a cohesive platform where madOS delivers AI-integrated OS infrastructure, Fractal-Mind provides cognitive memory architectures, Neuro Agent orchestrates multi-model AI workflows, and neuro-ollama enables GPU-accelerated local inference. Also the creator of madTrackers — a hardware/software solution for full-body VR tracking (SlimeVR/VRChat).
Core Competencies:
- AI & Knowledge Engineering — Multi-provider LLM orchestration, RAG pipelines, vector indexing (HNSW), fractal graph embeddings, cognitive memory systems
- Systems Programming — Custom Linux distributions, kernel tuning, hardware abstraction layers, Wayland compositing, memory optimization (ZRAM/zstd)
- Backend & APIs — High-performance Rust services (Axum), distributed architectures, REST/GraphQL APIs, real-time WebSocket systems
- Infrastructure & DevOps — Docker multi-service deployments, GitHub Actions CI/CD, GPU-accelerated containerization, automated ISO builds
- Embedded & Firmware — ESP32 IoT systems, VR/AR tracking (SlimeVR/madTrackers), nRF52 BLE bootloaders, Android native development
|
AI-orchestrated Arch Linux distribution optimized for resource-constrained environments (1.9GB+ RAM). Delivers adaptive GPU rendering with hardware acceleration fallback, ZRAM compression (zstd), high-fidelity audio support (192kHz/32-bit), and Sway Wayland compositor at ~67MB RAM. Fully automated ISO generation via GitHub Actions.
|
Production cognitive engine implementing RAPTOR-based fractal graphs with HNSW vector indexing over SurrealDB. Dual-phase architecture: Vigilia (real-time graph traversal) and REM (asynchronous consolidation). Features multi-user namespace isolation, multi-format ingestion (PDF/OCR), and multi-embedding support (Nomic, BGE, CLIP). Complete delivery — 12/12 roadmap features shipped.
|
Enterprise-grade AI programming assistant in Rust with intelligent multi-provider routing (Ollama, OpenAI, Anthropic, Groq). Implements AST-based code review with cyclomatic complexity analysis, LRU context caching (10× latency reduction), automatic error recovery with rollback, and comprehensive benchmarking suite with regression detection. 219+ tests, 100% pass rate.
|
Automated Git intelligence engine in Rust that synthesizes work summaries from repository activity. Processes commits, contributor patterns, and temporal trends with configurable time windows and multi-format export. LLM-powered natural language generation for executive summaries. Full CI/CD pipeline with comprehensive test coverage.
|
Tachikoma: Modular AI ecosystem combining GraphRAG memory, intelligent agents, and automatic model selection. Features SurrealDB for graph+vector storage with 11 relation types, multi-model routing (ministral-3b to qwen2.5-coder:14b), built-in tools (web search, command execution), and complete UI stack (React chat + admin dashboard + Z-Brain CLI).
|
IoT bridge for automatic WiFi connection on Android Auto using ESP32 microcontroller. Handles remote control and WiFi management as a hardware bridge between Android devices and vehicle infotainment systems. Demonstrates embedded systems expertise with resource-constrained hardware.
|
See all repositories at github.com/madkoding?tab=repositories
