Background / Problem
Motoru and EVPicker both showed me the same thing: people don’t just need “more info”, they need help thinking through car and finance decisions. EVs, tax rules and novated leases are full of jargon, edge cases and scary fine print. I wanted a space where I could experiment with different ways of helping people reason about those decisions without burning a big budget or building a full product.
Approach
What I built
I set up a single Firebase project as a sandbox and built a handful of tools inside it:
- I know my car – you enter a specific car and get an AI-generated report on running costs, features and suitability.
- Find me a car to lease – you answer a few lifestyle questions and get a shortlist of cars (any fuel type) that actually match how you live.
- EV novated lease calculator – simple – a quick, high-level estimate of tax savings and weekly costs for an FBT-exempt EV.
- EV novated lease calculator – detailed – a more serious calculator that models pre-tax/post-tax, on-road costs and out-of-pocket impact.
- AI EV Research Assistant – an AI-powered research tool that produces consistent, well-cited EV insight reports: key specs, ownership costs, public sentiment from real drivers, and summaries of pro reviews, all in one place.
Everything sits on Firebase, with AI APIs doing the heavy lifting for narrative explanations, comparisons and recommendations.
I built it alongside a novated lease consultant, using their feedback to keep the numbers sane and the language accurate enough to be useful but still human.
This was deliberately scrappy. Firebase gave me free hosting, auth and an easy way to keep everything in one place. There was a lot of vibe coding: wire something up quickly, see what comes out the other end, then tighten it once the shape is clear.
I focused on:
- Short, clear inputs – no giant forms.
- Solid, explicit calculations for the money side.
- AI used to explain, compare and suggest rather than “magic” everything.
Each tool started as a basic UI and a rough prompt. I walked through real scenarios with a novated lease consultant, then tightened the flows and prompts until the outputs felt both trustworthy and understandable.
Outcome / What I learned
The tools all work, and they became a live lab for questions like: How much can you safely offload to AI? Where do you need hard rules? How do you explain gnarly tax logic in plain English?
It also reinforced a pattern I keep bumping into: building is the easy part. Getting real people to use the tools, trust them, and come back is the harder, longer game. Working on this set forced me to think more about growth, onboarding and expectation-setting, not just “can I wire this up in Firebase?”
My role
I handled everything: product thinking, UX, interaction design, data and calculation logic, Firebase setup, AI integration and prompts, plus the ongoing tweaks with the novated lease consultant. The whole thing is still a working playground, and a base I can grow new EV and finance experiments from.
Closing note
Together, these experiments are where I try new ideas, kill weak ones quickly and sharpen how I think about systems. They’re not polished products, but they’ve had a big influence on the tools and experiences I’ve gone on to build, and there’s more to come. I’m always tinkering, learning new tools and looking for the next thing worth turning into something real.
Solution snapshots
A couple of screens from the Firebase toolkit.
