Portfolio · Szeged, Hungary

Szabolcs Veres‑Lakos

Mechanical engineer moving into embedded systems, IT and LLM tooling.

B.Sc. Mechanical EngineeringFirmware · Android · MCP github.com/vlszabolcs

1725.8601745.8701765.8801785.8901805.9001825.9101845.920

ITS-G5 channels, GHz. Channel 180 is the control channel where most V2X messages land.

About

I design railway weed-control equipment at G & G Kft. Along the way I took over the company's Autodesk Vault rollout and started automating CAD work with iLogic and Python. That side of the job turned out to be the part I enjoy most.

Now I spend my own time on hardware and software: porting radio firmware to new boards, building the app that talks to them, and learning how language models can drive real engineering tools.

Projects

V2X2MAP on T‑Dongle‑C5

In progress

A fork of V2X2MAP, an ITS-G5 receiver that listens to car-to-car (V2X) radio traffic and plots it on a map. I’m porting the firmware to the LilyGO T-Dongle-C5 and redesigning the Android companion app.

Hardware: LilyGO T-Dongle-C5 (ESP32-C5)
Radio: ITS-G5, 5.9 GHz
Links: USB serial, BLE
My part: Firmware port, Android UI

Building Firmware port and redesigned Android app
Planned Continuous RX over USB with no BLE time-sharing, non-blocking writes
Planned Standalone SD-card logging of raw frames, downloadable over USB or BLE
Planned GNSS module so the dongle can work as a mobile sniffer on its own
Planned Power-saving BLE modes and a split design: ESP32-C5 sniffer + ESP32-S3 comms

My fork (tdongle-c5)Upstream project

BrewPass coffee payment

Working

An RFID payment system for an office coffee machine. An ESP32 board takes the place of the machine’s coin validator: a user taps their RFID tag, the device checks and charges their credit in Firebase, then sends the same pulses a coin would. Credit is managed in an Android admin app, and a separate user app lets people buy a coffee or pass credit to a colleague from their phone.

Hardware: ESP32, RC522 RFID reader, SH1106 OLED
Machine: Saeco Royal Office (replaces NRI G-13.mft coin validator)
Backend: Firebase Realtime Database
Apps: Android admin app and user app (Kotlin, NFC)

Done Measured the coin validator interface with an oscilloscope: active-low coin and inhibit lines, 12 V supply
Done Card payment: credit check, charge and log in one database write, then the coin pulses to the machine
Done Rewrite with a hardware-independent core, native unit tests, FreeRTOS tasks on both cores and OTA updates
Done Admin app for users, top-ups and statistics
Done User app: NFC card binding, buying a coffee and transferring credit from the phone

Firmware (BrewPass)Android apps Demo video First version (rfidChecker)

LLM‑driven CAD with MCP

Learning

The Model Context Protocol lets a language model call tools through a standard interface. I’m studying freecad-mcp, an MCP server that lets an LLM create and edit FreeCAD models, to understand how such servers are built. My goal is to bring the same idea to the CAD automation I already do with iLogic and Python.

Protocol: Model Context Protocol
CAD: FreeCAD, Autodesk Inventor
Language: Python
Focus: Tool design for LLMs

neka-nat/freecad-mcp

Toolbox

Embedded & mobile

ESP32-C5 / ESP32-S3
Firmware porting, USB & BLE
Android app UI
RFID logging with external devs

Automation & LLM

Python basic
iLogic scripting
MCP servers learning
AI-assisted development

Engineering IT

Autodesk Vault ADMS admin
Autodesk Inventor
2nd-level user support
PDM / PLM concepts

Background

2023 – now
Mechanical Engineer, G & G Kft.
Product development for railway weed control. Vault PDM rollout and admin, iLogic automation, on-site engineering support.
2021 – 2023
Assembly Technician, G & G Kft.
Built the machines I now design.
2023
B.Sc. Mechanical Engineering
University of Szeged