About Me

I am part of a team that makes self-driving race cars. I work within the sub-department of the mapping team. When I started, I wrote a script which uses a modified incremental point fitting (ICP) algorithm to test the accuracy of our simultaneous localisation and mapping (SLAM) algorithms against a GPS reference.

Many iterative tests resulted in having to wait for github actions runners to start and stop. During busy hours waits upto 30 minutes were possible. Feeling this frustration I implemented a kubernetes management system which spins upto 8 containers split by cores to speed up development process. This is also known as a actions runner controller.

Now I'm working on optimising our SLAM algorithm for reliability for our upcoming competition in December 2026.

I was part of a team that made live telemetry software so engineers could see 300+ datapoints from the car live as it ran. I worked on various tasks, earlier I optimised dynamoDB calls to query preventing whole table scans and cutting latency and cost by 10x.

Later I worked on a feature called "math channels" which allowed engineers to create math equations on live data e.g. average speed of all four wheels. It was a unique system which calculated custom equations, identified equation dependents, and prevented cyclical dependencies between equations. It did this by representing equations as nodes in a graph and keeping the graph in a direct acyclic state. Each time an equation was added it checked for dependencies [using a dfs-cycle check] and could be stored in a dynamodb table for teamwide usage.

I built a custom CMS solution as part of contract work for a client. The frontend was designed with SEO optimisations in mind. It achieved 100 on Google's PageScore. The a backend was a webserver running Gunicorn and Python which took REST API calls to store to store entries using postgres. DNS routing and zone management were done our client's propriatery platform. The whole system was deployed on containers and updated through a github actions which uploaded the code and built it each time.

The biggest learning experience for me on this project was client management and team management. Setting correct expectations with clients and also managing internal team tasks and progress was a satisfying process.

Me and a group of four others built a robot designed to push other robots off a circular ring like Japanese sumo. The biggest challenge dealing with sensor noise and failures. Here I primarily wrote arduino code and tested for it's stability.