Rachel West

Cloud infrastructure for a federated learning platform

Volunteer contributor to UCL's Open Health CDI, a federated learning platform for health data built on Flower. I wrote the AWS infrastructure as code for the project's Level 0 training host in Terraform: provider and version pinning, adoption of the existing VPC through read only lookups rather than recreating it, an SSH security group, the GPU EC2 instance with a bootstrap script that installs the GPU drivers and Docker, and an Elastic IP so the host keeps its address across stop and start. I also fixed GPU detection in the test suite by following PCI device symlinks, and clarified the AWS host documentation.

Making the training backend dataset agnostic

Added a workload seam to the Flower server and client so the platform can train on datasets beyond the original PathMNIST benchmark: a workloads package with an active workload selector, wired through the server, client and container build.

Eight merged commits across infrastructure and backend, about 580 lines, all reviewed into the main project. onzelf/openhealth-cdi

Home automation and a home server

Home Assistant on a Raspberry Pi running lights, blinds, speakers and a wall dashboard by voice and remote, plus an always on Ubuntu Server box with network wide ad blocking, local language models and this site, managed over SSH.

Linux and infrastructure as code

Working towards RHCSA and Terraform certification, practised on real infrastructure at home and on the UCL project.