Descrição completa
The next ten years of AI will not be won in software. They will be won in the physical world: in factories, hospitals, kitchens, fields, and homes. The companies that own that data will own the century.
MicroAGI is building it. We are the data layer for physical AI.
Our models learn from data captured in the physical world, by hardware we design and build ourselves. As Embedded Engineer, you will own the firmware that runs on that hardware, from the first register write to the moment data leaves the device, so that what reaches our ML team is clean, synchronized, and trustworthy.
What You Will Do
Write and own the firmware for our data-capture hardware: sensors, motion-capture rigs, custom IoT setups, and prototype robots.
Build real-time data acquisition: drivers, DMA, buffering, and streaming from many sensors at once without dropping a sample.
Solve multi-sensor synchronization, calibration, and time-stamping so every data point lines up across devices and modalities.
Bring up new boards with our Electrical Engineer and turn fresh hardware into working devices fast.
Build the path from device to dataset: communication protocols, on-device logging, and handoff into our data pipeline.
Work shoulder to shoulder with ML engineers. Your firmware decides what their training data looks like, and a timing bug can quietly ruin months of data.
Make firmware reliable in the field: testing, OTA updates, error handling, and diagnostics that tell us what went wrong.
Requirements
Up to 5 years writing production firmware for embedded or IoT systems.
Strong C/C++, experience with an RTOS (FreeRTOS, Zephyr, or similar), and deep familiarity with microcontrollers (STM32, ESP32, NXP, or similar).
Hands-on experience with peripheral interfaces and protocols: SPI, I2C, UART, USB, BLE, CAN, or similar.
Experience integrating sensors in firmware: IMUs, cameras, depth sensors, force sensors, or similar.
A strong understanding of how firmware decisions affect data quality, especially for EMG/EEG sensing: ADC configuration, sample rate, jitter, filtering, and buffering. You know that a great model on a noisy signal still loses to a decent model on a clean one, and you write code accordingly.
Comfortable debugging at the hardware boundary with a logic analyzer, oscilloscope, and debugger.
Genuinely curious about ML: you read papers, you have trained at least one model end to end, and you understand that dataset quality beats model architecture choice.
High agency: you don't wait to be told what to do. You are comfortable in 0-to-1 ambiguity, where the spec is "figure out what the spec should be".
Discreet: you handle confidential hardware roadmaps and customer data with care.
Fluent in English. German is a plus.
Nice to Have
A background in robotics, motion capture, or wearable devices.
Experience with real-time or hard-deadline systems: audio, video sync, or robotic control loops.
Experience with embedded Linux, time-sync protocols (PTP or similar), or low-power design.
Familiarity with ML training pipelines. Even if you don't run them, you can read a training script.
Open-source firmware contributions.
Experience taking firmware from prototype to small-batch production: test fixtures, provisioning, and certification support.
Find Jobs in Germany on Arbeitnow
