Systems thinker with ADHD — I turn chaos into repeatable processes, whether it's construction governance, smart home automation, or cramming a Raspberry Pi into LEGO.

I hyperfocus on problems others ignore: "How do we audit 95.5% compliance in construction oversight?" or "Can this vintage LEGO set play actual Game Boy games?" Then I build systems that answer those questions.

My brain works differently, and I've learned to make it a strength:

Pattern recognition → I see connections between governance frameworks and Home Assistant automations
Hyperfocus loops → 15-hour deep dives into mmWave sensors or systemd services
Zero tolerance for repetition → If I do it twice, I automate it
Direct communication → No corporate speak, just "here's the problem, here's the fix"
Documentation obsession → Future-me will thank present-me for CLAUDE_CONTEXT.md

The projects below exist because ADHD doesn't let me leave problems half-solved.

Construction Oversight Framework (Reserva do Pinhal Works Commission)

• Problem: Homeowners association had no formal process for approving construction projects
• Solution: Built end-to-end governance system with 7 document templates, 36 technical criteria, automated protocol workflows
• Result: 95.5% compliance across multiple audit phases, 20-day SLA enforcement
• Tech: Python automation, systematic documentation, legal framework integration
• Status: Private (institutional governance), but ask me about process design

The hardest part wasn't the code — it was translating regulatory language into executable checklists.

Nabu Intelligence - Voice assistant that doesn't wake my mom

• Claude Sonnet 4 + Home Assistant Voice for natural language control
• Context-aware TTS with automatic volume reduction (quiet hours: 23h-7h, or manual "sleep mode")
• 30+ lights, 3 motorized curtains, A/C, 2 robot vacuums (named Caroline and Oswaldo)
• Maintenance alerts, smart scheduling, emergency backup systems
• Tech: ESPHome, YAML, Jinja2 templates, Anthropic API
• Repo: ha-nabu-intelligence-system (ready to publish)

Eldercare Monitoring (for my 83-year-old mother)

• Challenge: Monitor fall risk + CPAP compliance without cameras (privacy concerns)
• Stack: mmWave presence detection (LD2450/Aqara FP2), thermal imaging (MLX90640), audio anomaly detection
• Intelligence: Frigate + Hailo-8L AI for pose estimation (fall detection), contextual automation
• Privacy-first: No video recording, local processing only, thermal "blobs" instead of images
• Status: Active development, testing phase

ESP32 Irrigation System ("Lady Gaga")

• Standalone operation (doesn't depend on Home Assistant connectivity)
• Capacitive soil sensor (CS12) with pulse logic (25s ON / 120s OFF for deep watering)
• OLED display for local status, relay control with AC-rated protection
• Fallback safety: If HA dies, system continues autonomously
• Tech: ESP32-C3, ESPHome, fail-safe relay design

Home Energy Monitor

• Real-time power consumption tracking per circuit
• UPS monitoring with automatic alerts on power failure
• Integration with LoRa mesh for construction site monitoring

All projects follow the same philosophy: self-hosted, privacy-first, fail-safe defaults.

LEGO Game Boy Emulator (rpi-game-boy-lego)

• Converted LEGO set 72046 into functional Game Boy emulator
• Raspberry Pi Zero 2W + 2" LCD (ST7789), I2S audio (MAX98357A), 9 tactile buttons
• Challenge solved: PAM8403 amplifier burned out in <10s → migrated to I2S digital audio
• Systemd auto-start, safe shutdown via GPIO, 60 FPS GB/GBC emulation
• Status: 70% complete (display ✅, audio migration ⏳)
• Cost: R$462-631 (~$90-120 USD)
• Repo: Ready to publish with full build guide

Lessons learned: Always read datasheets before connecting components. PWM audio ≠ clean amplifier input.

Raspberry Pi Mini Console

• RetroPie multi-emulator with custom case design
• Focus on polish: boot animations, controller configs, kid-friendly UI

Tasmota Firmware Projects

• Flashing Sonoff devices for local control (no cloud dependencies)
• Custom MQTT integrations for X10 protocol smart home (for my brother)

Hardware Validation Methodology

• Rigorous component verification before soldering (datasheet-first approach)
• PCB design review, thermal calculations, failure mode analysis
• Why: Because I've burned enough components to know better

M5Stack Exploration

• Testing M5StickC Plus2, Cardputer, C6L for LoRa mesh networks
• Goal: Construction site monitoring + emergency backup comms
• Meshtastic community integration

Snapcast Multi-Room Audio

• Synchronized playback across Alexa devices
• Low-latency streaming with custom buffer tuning

Languages: Python 3.11+, JavaScript/TypeScript, YAML, Bash, C++ (ESP32)
Hardware: ESP32, Raspberry Pi, M5Stack, LoRa devices, mmWave sensors
Platforms: Home Assistant, ESPHome, RetroPie, Docker
Automation: Systemd services, MQTT, fail-safe state machines
AI/ML: Claude API integration, Frigate NVR, Hailo-8L accelerator
Docs: Markdown, technical writing, ADHD-friendly structure (lists > paragraphs)

Philosophy: Self-hosted where possible, cloud when necessary, always with local fallback.

Project Documentation
Every project gets:

• CLAUDE_CONTEXT.md - Full technical context for AI collaboration
• CHANGELOG.md - Semantic versioning (MAJOR.MINOR.PATCH)
• README.md - Human-readable quick start
• Incremental git commits (no batch commits at end)

ADHD-Adapted Workflow

1. Explore → Understand requirements, read existing code/docs
2. Plan → Architecture BEFORE implementation (avoid over-engineering)
3. Execute → TODO lists, commit incrementally, document decisions
4. Validate → Test edge cases, verify against spec

Critical rule: When in doubt, ask > assume and break things later.

What I'm good at:

• Translating vague requirements into structured systems
• Hardware validation (datasheets, schematics, "will this actually work?")
• ADHD-friendly documentation (concise, scannable, example-driven)
• Portuguese ↔ English technical translation
• Process design (governance, compliance, automation)

What I'm learning:

• AI/ML inference optimization (Hailo accelerators, TensorFlow Lite)
• LoRa mesh networking (Meshtastic protocol)
• Advanced ESPHome (custom components, lambda functions)

Languages: Portuguese (native), English (technical proficiency, still improving)

LEGO 72046 converted to functional Game Boy emulator. Complete build guide with component datasheets, troubleshooting, systemd integration.

Claude-powered voice assistant for Home Assistant. Context-aware responses, quiet hours automation, Portuguese BR support.

Autonomous irrigation system with fail-safe defaults. Works standalone if Home Assistant fails.

Universal documentation framework for technical projects. ADHD-optimized structure.

Location: São Paulo, Brazil
Interests: Governance automation, eldercare tech, retro computing, mesh networks, privacy-first IoT
Open to: Collaborations on Home Assistant integrations, hardware validation reviews, process automation

Not all projects are public yet:

• Governance work is institutional (privacy/legal constraints)
• Eldercare monitoring involves personal health data
• Some IoT projects are still in "prove it works" phase

I publish when systems are stable, documented, and genuinely useful to others — not just "it compiled once."

If you're building:

• Smart home systems that need to work when the internet dies
• Hardware projects where datasheets matter
• Governance frameworks for chaotic organizations
• Anything requiring ADHD-level attention to edge cases

Let's talk.

Built with hyperfocus in São Paulo 🇧🇷