Paul Mwangi

I’m an AI Research Engineer and full-stack developer specialized in end-to-end development of AI-driven applications and the evaluation of Large Language Models. I focus on making AI outputs accurate, safe, and dependable in real workflows.

I identify algorithmic bias, mitigate hallucinations, and audit complex codebases using Python, Java, and Dart. I also work with advanced NLP and LLM techniques like RLHF, Model Grounding, and LLM Benchmarking to test boundaries and improve reliability.

In my current role, I reduced unresolved discrepancies by 30% by investigating transaction variances with automated validation techniques, while maintaining 100% compliance with reporting deadlines. Previously, I developed Flutter mobile applications with robust Firebase backends and integrated Dialogflow with Telegram for conversational agents, including extensive flow testing and issue tracking.

I’m currently building my expertise through an MSc in Artificial Intelligence and hands-on projects spanning delivery workflows, inspection/maintenance tracking, merchandising monitoring, and a Smart Farm prototype combining mobile interaction with cloud data storage and sensor concepts.