Dedicated computer science professional with a Ph.D., skilled in Python, Java, and C++, seeking to leverage expertise in game development and machine learning.

PROJECTS

I am a seasoned professional in applied artificial intelligence and machine learning, and I hold a PhD in computer science from the University of Essex. My specialisation in memory-driven exploration and adaptive dynamic programming equips me uniquely to tackle complex AI challenges. My work includes the development of the Memory-Based Backpropagation Through Time (MBPTT) algorithm, which innovatively enhances decision-making in AI models by integrating memory processes.

With extensive experience in Python and cloud-native environments, I am adept at building production APIs and managing data delivery processes. My technical expertise is further enhanced by a robust background in software engineering, enabling me to contribute effectively throughout the software development lifecycle.

I am known for my collaborative approach and commitment to pushing technology's boundaries to achieve innovative solutions. My forward-thinking approach to technical problem-solving and strategic evaluation of architectural designs ensure that I am ready to impact your team significantly.

  • Developed the MBPTT algorithm, integrating memory with BPTT to handle long-term dependencies, enhancing decision-making quality and reinforcement signals in AI models.

  • Implemented a memory-augmented Q-learning algorithm for maze navigation, demonstrating the adaptability enhancements in dynamic environments and complex problem-solving.

  • Conducted research on the application of BPTT in complex, continuous spaces, broadening the scope from discrete environments to more intricate and dynamic settings.

  • Applied memory-augmented BPTT to a simulated food-seeking task, assessing the algorithm’s effectiveness in mimicking behaviours indicative of functional sentience and environmental mapping.

  • Evaluated various advanced memory models in conjunction with BPTT, analysing their benefits and limitations in reinforcement learning tasks and establishing a framework for comparison with traditional algorithms.

  • Undertook a detailed comparative study to evaluate traditional RL algorithms against memory-augmented BPTT agents, focusing on architectural aspects and the integration challenges of memory in RNNs.

Conferences:

Computing Conference 2023: attended and presented the paper “Finding Eulerian Tours in Mazes Using a Memory-Augmented Fixed Policy Function.” This presentation highlighted my work on applying memory-augmented algorithms for solving complex maze structures