Software Engineering
with Electronics.

Predict lap-time variation from stint length, air temperature and tyre compound, then check whether prediction error differed between team tiers.

I cleaned a 7,374-row dataset, encoded categorical features and trained a depth-limited DecisionTreeRegressor. I measured subgroup MAE with one unified model, then split the same training and test data by team tier and fitted a separate model to each group.

The specialised models produced much closer subgroup MAE values. Their overall error was slightly higher, showing that improved parity did not automatically mean improved predictive accuracy.

Design a robot that could detect maze walls, track its position and choose a route through an unknown 8 × 8 maze.

The prototype used an MBED LPC1768, seven active-low infrared sensors and two stepper motors driven through L297 controllers and ULN2003A arrays. The program grouped sensor inputs with BusIn, stored wall and distance grids, and used a sense-plan-act loop with flood-fill path planning.

Testing progressed from individual direction, sensor and wall-mapping functions to motor-driver and tunnel integration. The final demonstration confirmed wall detection and motor control; the report also records soldering faults and sensor-module substitutions encountered during integration.

Build a consistent website that remains clear and usable across desktop and smaller screens.

I used semantic HTML for the page structure, CSS for the visual system and responsive layout, and JavaScript for interface behaviour. The work focused on navigation, content hierarchy and reusable styling.

The finished project demonstrates my front-end fundamentals and gives me a codebase I can continue refining as my web-development skills grow.