Robotics Paper Replications
Replications of classic robotics algorithms to understand how they work. Here, RRT* keeps improving a route as its search tree grows. Other algorithms explore motion planning, feedback control and reinforcement learning.
Projects
TinyML Vision on Microcontrollers
A weekend project for quantisation. Handwritten-digit recognition on an STM32 microcontroller, with one model quantised to 8-bit. The network was deployed on the STM32F446RE using ONNX.
Hermes-Azure
Telemetry for cargo. Each shipment supports up to three monitoring nodes, sending readings over BLE to a base station that displays them to the driver and uploads them to a secure cloud ledger. More node-and-base-station groups can be added as needed.
This was a final project for an advanced embedded systems course. I built the Core2 interface in Zephyr to retrieve readings and alert the driver when conditions change.
Feedback4U
Rubric-aligned AI feedback for students and teachers. Teacher-provided reference materials give the model context via RAG. Feedback is organised by rubric criteria, so students can see where to focus their next revision.
This was a hackathon project in 2025. I integrated the application, retrieval service and structured feedback pipeline, and we ended up winning the hackathon.