VXT is a B2B startup that focuses on communication solutions for small- to medium-sized businesses globally. I built end-to-end integrations with various software—which included coordination with integration partners, API authorisation, database migrations, middleware and UI implementation. Additionally, I supervised a summer intern for the entire duration, and worked on several miscellaneous front-end redesigns, refactors, and rebuilds.

CreateBase was an EdTech startup that leveraged 3D simulations and programming tools to make tech education more accessible for schools. I worked on various tasks—from customer engagement, to product design, to implementation, but primarily focused on web development. HTML/CSS/JS and React were all self-taught to develop the website and app. We won first place in NZ for Red Bull Basement 2020, a global competition for university innovation, entrepreneurship, and startup ideas.

I was tasked with improving prediction accuracy of horizontal-axis washing machine run times. I collaborated with the product design team to understand the sampled signal data, then preprocessed, imputed, and feature engineered using Boruta random forest, XGBoost, and lasso regression. The extracted data was used to train a deep neural network. All programmes were written in R.

I applied deep reinforcement learning to robotic arms for improved efficiency in precision manufacturing. The agents were trained with a deep Q-learning policy using PyTorch and simulated in CoppeliaSim (written in Lua). I implemented a server-synchronous interface between client and server using the native simulation remote API.

Kessler is a student-run project aimed at launching a 1U CubeSat into LEO to verify the feasibility of electrodynamic tethers for rapid deorbiting. I wrote MATLAB code to calculate the power budget based on GMAT orbital simulations, and calculated accelerometer specs based on hardware, atmospheric drag, and plasma density in LEO. I also wrote the Orbital Debris Mitigation Plan for the mission.