Reality is full of uncertainty. As humans, we are adapted to this aspect of real life, and we can observe possible problems in plans and pivot in time, before failures become catastrophic. If we want autonomous systems deployed in real life environments, we expect them to behave just as well, if not better. We need them to be reliable. My work will focus on solving the problem of replanning with safety guarantees in the context of decentralized multi-agent systems, in the hopes that we will soon have autonomous fleets of vehicles safely deployed, as well as autonomous agents deployed in high risk situations where currently people are being sent, like search-and-rescue missions.
I chose to pursue a PhD with STAI due to my intention to expand my knowledge into the safety field of AI. The program gives us a significant amount of opportunities to learn and collaborate with others, and provides a structure that ensure me and my cohort are on track with our projects and responsibilities, while also enjoying the liberty of managing our own schedule for an unexpected work-life balance.
Masters Qualification: MEng Computing, Imperial College London
- Graduate Teaching Assistant and Undergraduate Research Assistant at Imperial College London
- Member of Technical Staff Intern at Nutanix in Cambridge, UK
- Game Development Teacher at ICHC, Romania
Publication: Transferring Multi-Agent Reinforcement Learning Policies for Autonomous Driving using Sim-to-Real: https://arxiv.org/abs/2203.11653