How I built ORBIX
ORBIX is a personal project by Deep Patel. I planned it and researched the engineering behind it, and I used AI coding assistants as my build tools.
Why I made ORBIX
I plan to study aerospace engineering. I wanted one place where I could look at real aircraft and launch vehicles and work through the engineering behind them: lift and drag, the rocket equation, orbital transfers, shock waves and entry heating. ORBIX is that place, and it is free for anyone to use.
My role
ORBIX was my idea, and I planned all of it: which aircraft and launch vehicles to cover, which calculators to build, how the pages fit together, and how the site should look. I researched the engineering behind the tools and directed every version, sending work back until it met the standard I set.
How I built it
I used AI coding assistants (Codex by OpenAI, then Claude Code by Anthropic) as my build tools. They wrote the code from my instructions, and they helped draft text and gather vehicle data, sources and image licenses, which I reviewed.
The project history is public on GitHub. Commits made with Claude Code are marked as co-authored by it.
Why I built it this way
AI is changing how software is written, and engineers in every field will work alongside these tools. I wanted to learn to use them well: to describe exactly what I want, to check what comes back, and to turn my own ideas and research into something other people can use.
What I learned
Building ORBIX taught me the basics of every aircraft and launch vehicle in it: what each one was designed to do and how its design serves that purpose.
The bigger lesson came from the Engineering Lab. Before any tool could be built, I had to research how it works: which equation it uses, what goes into it, and what the result means. Across every tool in the lab, that meant learning how lift, drag, thrust, gravity, shock waves and heating actually behave, and working out a way for each one to be calculated.
I also learned what it is like to run simulations: change one input, run it again, and see how an orbital transfer or a shock wave responds.
And I got real experience with how software gets made, from planning features to testing and fixing mistakes, along with using AI as a professional tool: giving precise instructions, checking what comes back, and not accepting something until it is right.
How the engineering is checked
The Verification page runs selected Engineering Lab calculations with the inputs of a published table or worked example and prints each result beside the published value. The results are not typed in: they come from the same functions the lab uses.
26 of 36 compared values fall within the rounding of the published figure. The other 10 are shown as they are, with a note on each.
How it is organized
The code is split into four layers. Vehicle records, mission presets and material data come first. Each calculator is a small function for one calculation, such as a Hohmann transfer, and analyses combine several calculators into one study. React components draw the pages and import from the layers above, and the equations stay out of them.
- An outline is a layer.
- An arrow is an import, from the imported layer to the importing one.
- The dashed line is the presentation boundary, with React below it.
Two GitHub Actions workflows run the checks below. The first six run together as npm run validate.
Validate workflow: every push to main and every pull request
| Check | Command | Tool |
|---|---|---|
FormattingPrettiernpm run format:check | npm run format:check | Prettier |
Lint, zero warnings allowedESLintnpm run lint | npm run lint | ESLint |
Design rules: no file may add raw colors, gradients, shadows or other banned styles beyond its recorded baselineNode scriptnpm run check:design | npm run check:design | Node script |
Type checkTypeScriptnpm run typecheck | npm run typecheck | TypeScript |
Unit tests for calculators, analyses and componentsVitestnpm run test | npm run test | Vitest |
Production buildNext.jsnpm run build | npm run build | Next.js |
Browser tests workflow
| Check | Command | Tool |
|---|---|---|
Smoke and accessibility tests at desktop, tablet and mobile sizes, on every push to main and every pull requestPlaywright, Chromiumnpx playwright test tests/ | npx playwright test tests/ | Playwright, Chromium |
Visual regression screenshots, only when the workflow is run by handPlaywright, Chromiumnpm run test:visual | npm run test:visual | Playwright, Chromium |
Both workflows can also be started by hand from GitHub Actions.
Project facts
- Started
- August 2026
- Built with
- Next.js, React, TypeScript
- Contact
- deep.patel4000@gmail.com