- Collaborated with a team of 8 on a research project centered on Tiny Machine Learning (TinyML) for IoT devices
- On-device computation through effective memory management to enhance performance on a microcontroller
- Achieved above 80% model compression through selective quantization techniques while maintaining accuracy
- Deployed onto diverse SoC including nrf5340, BCM2710A1, and RP2040 with an accuracy loss of less than 2
- Innovated ML adaptations for object detection and sensor data analytics, minimizing dependency on the cloud
- Tech Stack: Python, TensorFlow, C/C++, Linux (Ubuntu), TensorFlow Lite, Zephyr RTOS, arm-gcc compilers