EIP ERO Portal, July '23 - Present

Working on back-office application to process applications for postal and proxy votes. Highlights includes:

• Ownership of complex feature from initial design through implementation and demos to the client. Worked with designers and produced high-quality implementation that has been bug-free since release.
• Led several knowledge-sharing sessions on topics across our tech stack; supported new starters in the team.
• Brought in close to deadlines to get features finished on several occasions while balancing the priorities of my own work-streams.

Kotlin, Typescript, React, Terraform, MySQL, AWS

EIP Citizen-Facing Services, Feb '23 - July '23

Worked on citizen-facing websites for postal and proxy votes, as well as various adjacent services. Highlights include:

• Converted critical set of AWS lambdas to Typescript, identifying and fixing long-standing issues in the process.
• Investigated and implemented solution to limitation with existing approach to storing form values. Ensured approach is robust to future changes.
• High impact team-member, noted by managers for finding tricky bugs, identifying design inconsistencies or improvements and for being able to pick up new technologies quickly.

Javascript, Java, Scala, Terraform, AWS

Training & internal apps, Sept '22 - Feb '23

Worked on various projects during training including:

• An HR tool tailored towards Romanian companies.
• Scripts to export our hiring database into a format accepted by a new third party tool.
• An app to help learn employees' names.

React, Python, Ruby

Masters Dissertation

Multi-Agent Reinforcement Learning for Search and Rescue

Led a team of 6 on an investigation over the course of the year into the utility of a variety of multi-agent reinforcement learning algorithms in the field of search and rescue. We developed a custom sandbox environment to model potential scenarios at a high level and worked with RLlib — a scalable deep reinforcement learning library — to trial a variety of disaster scenarios, using heterogeneous agent types with different abilities. We concluded that there is potential for these algorithms to guide agents at a high level on the tasks to carry out — exploration, search, rescue or recovery — but that control at a micro level will still likely require human intervention at a minimum.

Dissertation

Trajectory Planning for Autonomous Racecars

Working with a small-scale wheeled robot, I used computer vision to detect the driveable surface, plan a trajectory and follow it. Due to issues with setting up a binocular vision system, I used monocular vision with planar transformations to carry this out.

Modules

• Image Processing
• Machine Learning, Neural Networks, Data Science & Analytics
• Game Theory
• Project Management for Computer Science
• Databases
• Fault-Tolerant Systems
• Algorithms, other Computer Science Fundamentals

Activities

Pre-Us

• Maths (D1/A*)
• Further Maths (D2/A*)
• Physics (D1/A*)
• Economics (D1/A*)

Activities

Expedition society, prefect, front-desk cellist in Symphony Orchestra, football, community volunteering.