About

**Project type:** One-time project, with potential follow-on work **Scope:** Firmware + basic protocol design / integration, committing to existing code base **Hardware:** Seeed XIAO ESP32-S3 Sense (camera) + Wi-Fi HaLow (802.11ah) expansion --- #### Overview I’m building a **low-power vision node** that can sit in the field, watch a scene, and periodically report **compact statistics**, not full images. Example use cases: * Tracking color blobs and patterns (e.g., traffic light states over time) * OCR / digit recognition from fixed regions on a screen * Gauge / dial reading (analog needle position → numeric value) The node uses: * **ESP32-S3** (XIAO ESP32-S3 Sense with camera) for image capture and edge processing * A **Wi-Fi HaLow (802.11ah)** expansion board for long-range, low-power IP connectivity We need an embedded engineer to design and implement the application portion of the firmware. We have an existing code base that handles camera, HaLow, settings and other features. --- #### The challenge * The device must spend **most of its time in low-power mode**, waking up periodically to: * Grab a frame from the camera * Run one or more *simple detection pipelines* (color blob / OCR / gauge) * Send a small JSON/binary payload over HaLow (e.g., “gauge=137.2”, “light=GREEN”) * The same physical node needs to be **reconfigurable in the field** to act as: * A color-blob observer, **or** * An OCR node, **or** * A gauge/dial reader * For setup and tuning, the device must support a **“calibration/streaming mode”**: * Normally: low-power “snapshot” mode (no streaming) * On command via radio: temporarily stay awake, stream camera frames (e.g., MJPEG or similar) so a remote UI can draw ROIs and tweak parameters * After a timeout or explicit command, go back to low-power mode #### Ideal experience / skills * Strong **C/C++** on microcontrollers * Hands-on experience with **ESP32 / ESP32-S3** (camera drivers, FreeRTOS, power management) * Familiarity with **Wi-Fi HaLow (802.11ah)** or at least working with external RF modules / AT command stacks * Low-power design: deep sleep/light sleep, duty cycling radios, minimizing active time * Basic image processing on embedded systems: * Working with low-res grayscale / RGB frames * ROI extraction, thresholding, simple feature extraction * Experience with **TensorFlow Lite Micro / ESP-DL** is a plus, but not strictly required if classical CV methods suffice * Experience implementing **OTA updates** and resilient embedded networking is a bonus --- #### What to include in your proposal * A short description of **similar projects** you’ve done (ESP32, vision, low power, or long-range wireless) * Your preferred stack (**ESP-IDF vs Arduino/PlatformIO**, streaming approach, etc.) * Any thoughts on: * How you’d structure the pipeline / mode system * How you’d implement the streaming vs low-power state machine If this sounds interesting and you’re comfortable working fairly close to the hardware (with real constraints on power, bandwidth, and RAM), please apply. More information available upon request, feel free to ask questions.