Where AI meets hardware
Before AI, I worked on physical devices: an air-conditioning monitoring system in a university lab, CAN-bus data acquisition from energy-storage batteries, a glucose meter prototype on Arduino, and the telemetry ground station for a CanSat.
It is a concrete advantage: AI models are only as good as the data they learn from, and good data is designed starting from the sensor. I can cover the whole chain, from the firmware to the model that forecasts on the collected data.
What I build
- Sensor data acquisition: reading, sampling, calibration and validation of values.
- C and C++ firmware for microcontrollers such as Arduino, with buses and protocols such as I²C and serial.
- CAN bus and energy storage: reading frames, decoding them and analysing battery performance.
- Telemetry: packets with delimiters and checksums, ground stations and logging of every session.
- Test automation and certificates: acceptance reports and conformity documents generated from test data.
- Dashboards and AI models in Python on the collected data: anomalies, predictive maintenance, forecasting.
How we work
- Signal requirementsWhich quantities you need, how often and how precisely, and what happens when a sensor fails or disconnects.
- Firmware and communicationReading sensors, timestamps, plausibility checks and packets with delimiters and checksums, over serial, I²C, CAN or radio.
- LoggingEvery session saved with raw data and metadata: the basis for debugging, acceptance tests and future models.
- Processing and visualisationFilters, conversions and derived quantities in Python, with a clear dashboard for whoever uses the device.
- Testing the whole chainKnown values injected from the sensor to the screen, to find errors where they start.
Real projects
Sensor data acquisition, processing, logging and visualization for an IoT air-conditioning monitoring device.
CAN-bus reader & automated battery test certification, written for Voltfang's incoming-product test pipeline.
Arduino + INA219 current sensor + I²C LCD: strip detection, peak timing and a calibration line from shunt voltage to mg/dL. A study prototype, open source.
Serial telemetry with fixed 100-byte packets and start/end markers, live altimeter, compass and artificial horizon in JavaFX, one CSV log per session.
Technology
Further reading
Before AI models there is data, and good data is designed starting from the sensor. The method I use, from firmware to logging to the dashboard, learned on monitoring devices and energy-storage systems.
Notes from the Voltfang internship: parsing CAN frames, decoding DBCs, automating conformity certificates.
In 2020 I built a digital glucose meter prototype with an Arduino, an INA219 current sensor and an LCD. How it works, where the code was fragile, and why a prototype is not a medical device.
In 2018 I wrote the telemetry ground station for a CanSat project for the AIAA competition: serial packets with checksums, flight instruments in JavaFX and a log file for every session. What I would do the same, and what not.
Frequently asked questions
Do you also design the electronics?
My work focuses on firmware, data acquisition and the software around the device; in prototypes I have integrated ready-made sensors and modules, such as I²C current sensors and displays. Designing complex circuit boards should be done together with a hardware designer.
Can our machines be connected to a monitoring system?
Often, yes: many machines already expose data over CAN bus, serial lines or other protocols. The first step is understanding which data is available, how often and how reliably.
Is sensor data useful for predictive maintenance?
Yes, if it is collected continuously and with the right references, for example when a failure happened and which interventions were made. Designing the acquisition well is the first step of any predictive-maintenance project.
Do you work on battery devices and energy storage?
Yes: at Voltfang I built the package that reads CAN data from energy-storage batteries, and the automatic generation of acceptance reports and conformity certificates.