At Oculotronics, I contribute to a five-camera automated inspection and rejection system that connects STM32 sensor and actuator control with computer vision. I developed firmware logic that triggers image acquisition when a photoelectric sensor detects a container and relays inspection results over UART.
I implemented timed GPIO and solenoid pulses to reject defective containers, and developed open-loop tracking to synchronize actuator timing with conveyor movement. I also extended serial telemetry from two sensors to five, updating the firmware data format and Python parsing.
For the inspection pipeline, I worked on OpenCV image preprocessing, including stitching and cylindrical projection for 360° panoramas. I prepared defect-inspection datasets and supported YOLO model training, validation, and per-class error analysis.
At Tech Mahindra, I developed register-level peripheral drivers and reusable SPI EEPROM modules for STM32 boards. I also used FreeRTOS tasks and semaphores to coordinate shared SPI access, and debugged firmware with JTAG/SWD and an oscilloscope.
