Coffee Kitty

Inspiration

Coffee Kitty was started at the Medo Vibe Coding Hackathon held with AI builders in Singapore last month. The original called Kopi Kitty build with my team there was selected as top 20 finalists and we presented it at the hackathon! It was inspired when my team and I realised that Singaporeans depend heavily on coffee in their everyday lives, so this is to optimise efficiency and health at the same time.

What It Does

It is a personalised caffeine scheduling app that uses pharmacokinetic modelling to generate an optimised coffee (or matcha) plan for your study session or exam day.

You tell it:

How long you slept
Your study window (or exact exam time)
Your profile — age, weight, caffeine tolerance, workload
Any busy time blocks to avoid (lectures, lunch, etc.)

It gives you:

A scientifically-modelled alertness curve over time
Exact coffee timings with ☕ plotted directly on the graph\
A health tracker with caffeine warnings
A printable Coffee Kitty Café receipt with your total spend and peak alertness
Post-exam recovery advice
A little animated cat for moral support

The Science Behind It

The core engine implements the Bateman equation — the standard single-compartment pharmacokinetic model for caffeine absorption and elimination:

$$C(t) = \frac{F \cdot D \cdot k_a}{V \cdot (k_a - k_e)} \left( e^{-k_e t} - e^{-k_a t} \right)$$

Where:

$ k_a = 4.5 \, \text{hr}^{-1} $ — absorption rate constant
$ k_e = \frac{\ln 2}{5} \approx 0.139 \, \text{hr}^{-1} $ — elimination rate (half-life ≈ 5 hours)
$ V = 0.5 \times \text{weight (kg)} $ — volume of distribution
$ F = 1.0 $ — oral bioavailability

A greedy optimizer places doses ~45 minutes before key alertness windows, respects your daily caffeine cap (200mg for low tolerance, 400mg otherwise), and enforces a hard 6-hour pre-sleep cutoff.

How I Built It

For this revamp, I brought the full spec to MeDo in a single structured prompt — the pharmacokinetic equations, the kopitiam menu, the Chart.js custom plugin for rendering emoji on the curve, the profile system, the receipt generator, all of it. MeDo handled the full-stack generation, producing a clean single-file HTML/JS application.

The thing I found most impressive: MeDo correctly implemented pharmacokinetic superposition — multiple coffee doses stacking their concentration curves the way they actually do in your bloodstream, without me writing a single line of code myself.

Challenges

Accurately calculating caffeine and alertness levels

Forming a team as a non-coder at the original hackathon was its own challenge — and honestly the most memorable one. Walking up to strangers and making friends to vibe code together was initially a big task.

What I Learned

As someone who does not code, I learned that Medo is bringing coding to the general audience by enabling us to use AI.

What's Next

Wearable integration (heart rate as a real-time alertness proxy)
Personalised half-life calibration (some people metabolise caffeine 2–3× slower)
Group study mode — sync coffee breaks with friends
Actual cafe partnerships for in-app ordering