I build embedded firmware
and full-stack web
applications.

I implemented BCH4 and BCH8 error-correcting algorithms in the NAND flash drivers to maintain data integrity despite wear and bit-flips. I also debugged and fixed issues in the FATFS file system, improving file handling and storage stability on the production devices.

Development of progressive web applications with Next.js and PostgreSQL, with a focus on performance and scalability.

A complete project covering both firmware and desktop software. The firmware sends button and analog-pedal events over USB as formatted messages such as (BTN1_PRESSED) and (THROTTLE:12345). A Windows application maps these to either keyboard input or a virtual joystick through the vJoy driver, and handles port selection, mode switching, analog pedals, a developer console, and automatic configuration saving. ESP32-C3 firmware and a flashing kit are planned.

A work-in-progress micro-operating system for embedded boards, currently targeting the ESP32-C3. I developed it to deepen my understanding of operating-system design and to provide a reusable base for my own projects. It includes a lightweight kernel, a cooperative and priority-based scheduler, a task manager, a hardware abstraction layer for GPIO, timers and peripherals, an IPC system, timer utilities, and a serial command-line shell for debugging over USB-JTAG. It is designed to be modular and portable across microcontroller families. Written in C and under active development.

I developed the low-level drivers for the USB-CDC and USB-MSC device classes and the One-Wire bus, and integrated a local file system for efficient on-device data storage. I later returned as a freelancer in 2026 to resolve edge cases in the CDC driver.

I automated the firmware flashing process for gauges on the production line, removing a manual bottleneck and making device programming fast and repeatable.

A real-time navigation system that converts LIDAR scans into occupancy grids, detects obstacles, and plans a path that adapts as the environment changes. It performed autonomous navigation through complex routes with dynamic obstacle handling, demonstrating that low-cost LIDAR is sufficient for accurate navigation.

A face detection and recognition system built in Python and OpenCV and connected to a student database. It logged recognised students automatically with timestamps to prevent proxy attendance, and achieved over 95% accuracy under controlled conditions.