James Watson Homer

Computer Science Student & Software Engineer

Aiming for honours and a PhD in AI. Formerly Senior Modelling and Control Engineer with the USYD Rocketry Team.

Experience & Education

2024 - Present

University of Sydney

Bachelor of Advanced Computing (Computer Science)

Studying as a Dalyell Scholar, graduating June 2028. I'm interested in all aspects of my learning, planning on completing honours and a PhD in AI.

USYD Rocketry Team

Senior Modelling and Control Engineer · Aug 2025 – Aug 2026

Led the integration of airbrakes into Ironbark, the team's trajectory simulator, and mentored two junior members. Ironbark's oxidizer fill estimation and control algorithm fitting helped the team win category first place at IREC 2026.

Ground Control Software Engineer · Sep 2024 – Aug 2025

Created CLI data management tools with FTP to safely handle telemetry data, designed laser-cut panels, and aided in the design of the launch controller. The team won first place at IREC 2025 in Midland, Texas, against 142 university teams from 22 countries.

2021 - 2024

The International High School of the Gothenburg Region

International Baccalaureate Diploma

Focused on STEM subjects while maintaining broad interests in the humanities, particularly History.

2019 - 2021

Self-Directed Learning & Freelance Work

Programming & Client Projects

Learnt Java through Minecraft plugin development, expanded skills to C++, JavaScript, and Python while taking on Minecraft plugin freelance projects and developing client management skills.

2006

Born in Sydney, Australia 🇦🇺

Technical Skills

skills.sh
james@homer:~$
Try commands: help, about, ls, cd [section]

Projects

Jarv

Jarv is a scriptable, interactive AI agent TUI in Python that talks to several model providers directly over httpx.

  • Agent tools let it browse the web, edit codebases, and change computer settings, with automatic cost tracking built in.
  • Discovers models and pricing live from each provider's /models endpoint, so listings stay current without updating Jarv.
  • An automatic auditor reviews risky commands, reducing the effectiveness of prompt injection attacks and unwanted behaviour.

Ironbark

Ironbark is the USYD Rocketry Team's 6-DOF trajectory solver. I'm its lead developer and maintainer.

  • Implemented airbrake simulation and a modular framework for control algorithms, built for the team's first actively controlled rocket.
  • Migrated the aerodynamics from panel solver techniques to Gaussian Processes (scikit-learn) fitted on CFD-derived data.
  • That migration significantly improved sample efficiency, simulation accuracy, and interpolation simplicity.

Interests

AI & LLMs

Hands-on experience with the OpenAI (Usage Tier 4), Gemini, and Claude APIs. Fine-tuned large language models and built mapbench, a benchmark experiment evaluating spatial reasoning in language models. Actively learning PyTorch and reading papers on the way to honours and a PhD in AI.

Rocketry

Two years of competition rocketry with the USYD ROcketry Team, first in gorund control software then modelling and control. Most interested in simulating reality as well as building the pipelines and software to enable such things efficiently and in an automated fashion.

LLM Applications

Built Slides2Textbook, an open-source CLI tool that uses LLMs to convert lecture slides and transcripts into coherent, well-formatted textbook chapters. I'm interested in the very many current and future practical applications of language models.