Talha’s project centred on building a fully driverless racing simulation in Unreal Engine 5, combining advanced game development with practical artificial intelligence. He trained multiple AI agents to navigate racetracks independently, compete against one another, and race against a human-controlled car.

Across three terms, Talha engineered a complete learning pipeline of his own design. He began by creating a custom matrix API and a regression-based neural network API capable of training models from any 1D dataset. Building on this foundation, he moved into reinforcement learning, first producing a Deep Q-Learning maze-solving agent and then extending the same principles to control a virtual vehicle.

To support this, Talha explored spline-based racetrack systems, integrating them into a robust training environment where each agent could learn through trial, error, and refinement. The final result was a sophisticated driverless racing system that showcased both technical depth and creativity.

This project captures Talha’s interest in artificial intelligence at a university-level standard and now forms a standout centrepiece in his applications for Computer Science and Artificial Intelligence.