Software · Embedded · Systems Developer

C/C++ · Python · TypeScript · Kotlin · Rust · ESP-IDF · Linux · Android

I build practical systems across embedded firmware, developer infrastructure, backend/web products, mobile applications, applied ML and hardware tooling.

I am most interested in projects where software has to deal with real constraints: limited memory, hardware state, wireless protocols, process failures, unreliable networks, persistent data and long-running operation. I prefer explicit contracts, diagnostics and testable boundaries over hidden behavior.

Public showcase for my larger private ESP32-S3 firmware platform. The system combines a node-based automation runtime, sensor integrations, BLE/MQTT connectivity, microSD history/archive, an embedded web interface and a Clay-rendered device UI across multiple display targets.

The private firmware repository uses host-side Nodeflow tests, ASan/UBSan runs, coverage, static analysis, firmware architecture guardrails, memory-profile checks and load-testing tooling. The public repository exposes the product/architecture without publishing the full firmware source.

Stack: ESP32-S3 · C/C++ · PlatformIO · FreeRTOS · BLE · MQTT · microSD · Clay · TypeScript/Svelte

Public showcase: https://michalmatu.github.io/growclip/

Deterministic local execution infrastructure for AI-assisted development. A planner defines the work; local-agent executes bounded tasks on the local machine and publishes machine-readable evidence of what actually happened.

The current implementation includes multi-repository scheduling, isolated workspaces, OS execution leases, process-group lifecycle management, command/no-output/RSS watchdogs, durable task claims and result spooling, crash recovery, bounded Git retry behavior, self-update safeguards and macOS launchd deployment.

Engineering signals: Linux/macOS process semantics · concurrency · durable state · failure recovery · resource limits · Git orchestration · unit/integration tests · macOS smoke tests

Stack: Python · Git · subprocess/process groups · threads · file locking · launchd · GitHub Actions

ESP32-S3 controller research project combining a portable C++ control library, deterministic safety supervision, generated TinyML inference, Python model/simulation tooling and browser-based hardware/chamber configurators.

The portable controller is intentionally separated from ESP-IDF/Arduino concerns. CI verifies Python and ML tooling, portable C++ host tests, clang-tidy, an ESP-IDF firmware build plus clang-check, and the React/TypeScript frontend independently.

Stack: ESP32-S3 · ESP-IDF · C++17 · CMake/CTest · TinyML · Python · React · Three.js

3D chamber configurator: https://michalmatu.github.io/growbox-ml-controller/chamber-3d

ESP32-S3 sensor/display platform that combines environmental sensing, BLE scanning, Wi-Fi CSI motion sensing, USB HID, local logging, HTTPS/JWT security, notifications and a SvelteKit control interface.

The project includes native host-test environments and coverage support, API/architecture documentation, flash coredumps and explicit PSRAM/TLS memory tuning for a resource-constrained device.

Stack: ESP32-S3 · C/C++ · PlatformIO · BLE · Wi-Fi CSI · TinyUSB · SCD41 · LittleFS · SvelteKit

A map-based application for documenting long-standing vehicles in public space. It combines field-photo ingestion and review, SQLite-backed state, privacy-safe image derivatives, historical WMS imagery, cadastral data, PDF report generation, administration and backup/restore tooling.

The repository has separated HTTP/core layers, a broad Python test suite, frontend checks, architecture/data diagnostics, CI, smoke tests and an end-to-end upload → review → map → PDF flow.

Stack: Python · SQLite · Leaflet · WMS/GIS · Pillow/OpenCV · ReportLab · Cloudflare Tunnel · Restic

Live: https://ilestoi.pl

Visual place-discovery and editorial map platform built around a place-centric data model. The project includes a FastAPI/SQLModel backend, Alembic migrations, media/content pipelines, moderation-oriented services and a React/Leaflet frontend.

The codebase is split into API, core, database, models, schemas, serializers, services and tests. CI runs a common quality gate with backend/frontend validation and Playwright-based browser checks.

Stack: Python · FastAPI · SQLModel · Alembic · React · TypeScript · Leaflet · Vite · Playwright

Live: https://photomap.pl

Real-time IP-camera monitoring and evidence recording with YOLO-based dog / person detection. The pipeline supports ROI-aware inference, configurable runtime detection profiles, adaptive idle/active detection cadence, full-frame recording and automatic MPS → CPU fallback.

Stack: Python · OpenCV · YOLO/Ultralytics · Flask · MPS · OpenVINO · JavaScript

Modular Android application for observing Bluetooth/BLE devices, maintaining a watchlist and presenting evidence/confidence around detections. It includes foreground scanning, Room-backed data, BLE decoder modules, Compose features and instrumented tests.

Stack: Kotlin · Jetpack Compose · BLE/Bluetooth · Room · Hilt · Android services

MediaPipe-based Android project for on-device gesture recognition and gesture-to-action mapping, with explicit lifecycle handling, bounded inference behavior, privacy-oriented offline operation and an Android quality gate in CI.

Stack: Kotlin · Android · MediaPipe · CameraX · Gradle

• Local Climate Link — local BLE/BTHome thermometer → Shelly automation without a cloud service or central hub.
• Miauudio — Android-first ambient audio mixer built on a web/Capacitor stack with native Android media integration.
• Arduino MQ — ATmega328P environmental/gas-sensor project with explicit safety limitations and local display/UI.
• LilyGo 4.7 — ESP32-S3 e-paper/touch integration with LVGL and a separated hardware/UI structure.

I use AI coding agents extensively as part of my normal development workflow. I treat them as engineering tools rather than as a substitute for verification: I define constraints and architecture, review changes, debug integration failures and validate behavior with the strongest checks that make sense for the project.

Depending on the repository, that includes:

• host/unit/integration tests and hardware smoke tests,
• compiler and static-analysis gates,
• ASan/UBSan, coverage and memory/resource diagnostics,
• schema/API/data-contract validation,
• reproducible CI builds and end-to-end flows,
• physical-device validation for firmware and hardware-facing work.

The local-agent project is also part of this workflow: it provides deterministic, bounded execution of agent-planned tasks across my local repositories.

Languages: C · C++ · Python · TypeScript/JavaScript · Kotlin · Rust

Embedded: ESP-IDF · PlatformIO · FreeRTOS · BLE · Zigbee · Wi-Fi · MQTT · USB/TinyUSB · I²C/SPI/UART

Software: FastAPI · SQLite/SQLModel · React · SvelteKit · Android/Compose · Linux/systemd · Git automation

ML / vision: TinyML · YOLO · OpenCV · MediaPipe · simulation/calibration tooling

Hardware: KiCad · SKiDL · PCB automation · sensors · displays · power management

Quality: GitHub Actions · host tests · integration/e2e tests · clang-tidy · cppcheck · Ruff · sanitizers · coverage · static analysis