I am a Python engineer with a passion for AI, AI-driven development, concurrency, and image processing.

I also have extensive experience in developing hardware and software prototypes, specializing in C/C++ for ESP32 & ESP-IDF. I enjoy working with communication protocols like CAN Bus, UART, I2C, and SPI, and integrating various sensors and devices such as GPS, LiDAR, etc.

Here is a curated list of projects I have designed and built:

A lightweight, educational framework for building LLM agent workflows using a declarative state graph with deterministic Bulk Synchronous Parallel (BSP) execution.

This project demonstrates a deep understanding of complex AI orchestration by solving the notorious "black box" debugging problem common in modern agent frameworks. By enforcing strict immutability and predictable execution cycles, it provides a robust, testable, and transparent architecture.

• Deterministic Execution: Uses a fixed Compute → Barrier → Apply cycle (BSP model) for predictable and reproducible agent traces.
• Immutable State: State is a frozen dataclass with explicit patches and typed reducers, preventing parallel nodes from silently overwriting each other.
• Auto-generated GUI: Automatically generates a web-based UI featuring live state graph visualization, runtime LLM parameter configuration, and real-time execution tracking.

A shared Cursor IDE configuration library containing curated .mdc rules and agent skills that enforce consistent coding standards and best practices across projects.

• Agentic Project Management (APM): Implements a structured workflow dividing AI tasks into roles: Planner (high-context reasoning models), Coder (fast implementation models), and Human (Tech Lead oversight).
• Automated Quality: Enforces strict boundaries, Clean Code, SOLID principles, automated linting/formatting (Ruff, ESLint), and strict type checking.

A Python tool that adds transparent file encryption to any git repository — selected files stay readable locally but are encrypted before they reach the remote repository.

• How it works: Leverages git's native clean/smudge filter mechanism using age encryption — you use git exactly as before, requiring no workflow changes.
• Key Features: Pattern-based file selection, multi-recipient collaboration, ensures only encrypted data is stored in the git history (plaintext never leaves your machine), and an interactive setup wizard with multi-language support (EN, DE, FR, ES, CS, PL).

Where Threads and Processes Sing Together. A comprehensive Python framework for managing concurrency in applications involving multiple processes and threads.

• Real-world Origin: Developed for high-parallelism sensor data logging (GPS, CAN bus, rangefinder) on Raspberry Pi 4b aboard drones at Zuri.com, with the kind permission of Michal Illich.
• Key Features: Supports multi-threading/multi-processing, advanced signal management (standard & user-defined signals), state persistence via files, and Tkinter-based GUI integration for process control.

Affordable image processing for assistive technologies. A concept and library designed to extract facial and body pose data, providing text outputs via an API.

• Live Demo / Info: Visit the Astriot Website
• Purpose: Enables affordable, lightweight image processing for accessibility and other sensor-driven applications.

A dedicated GitHub organization containing a collection of modular, production-ready CAN libraries for ESP32 platforms using ESP-IDF.

• can-multibackend-idf: Multi-Backend CAN Example Suite. Seamlessly switch between built-in TWAI and external MCP25xxx controllers over SPI.
• twai-idf-can: Simplified high-level adapter for ESP32's built-in TWAI (CAN) controller with automatic error recovery.
• mcp25xxx-multi-idf-can: Extended CAN driver supporting multiple MCP2515 controllers over SPI.
• examples-utils-idf-can: Common utility functions shared across the CAN examples.