Experiment 01 · Machine Learning in Bio-Mechatronics & Bioinformatics
Don’t just train a model on someone else’s dataset. Collect your own biological data, build your own dataset, and train your own model.
Put your fingertip on a MAX30100 light sensor. Record yourself resting, moving and recovering. Teach a model to tell the three apart, then put the model on OrbitCore so the board predicts by itself and blinks the answer.
About 40 minutes with a board. Works in Chrome or Edge on a laptop. Your recording stays in this browser tab.

The whole journey
Select a step to jump to it. Steps open one at a time as you finish the one before.
The MAX30100 shines infrared and red light into your fingertip, 50 times a second.
Each heartbeat pushes a little more blood into the finger, and blood absorbs light. The light coming back dips with every beat.
From those dips you build a dataset, train a model, and copy the model onto OrbitCore. It then blinks 1, 2 or 3 times for Baseline, Activity or Recovery.
The question you will answer
“Does giving the model more information improve classification?”
You will train models that see only the heart rate, all the infrared features, or infrared and red light together. Then you compare their scores on your own data. The answer is not decided in advance; it depends on what you record.
What you need
No board today? Use the simulated signals to follow every step.
1Physical Vibe Coding
A sensor only makes a tiny voltage. The board needs a small program to read it. Say what you want in your own words, and the code is written for you.
Ask for the sensor, then change it step by step: which lights to send, how bright.
Nothing yet. Type a request below or pick one of the examples.
In simple words
The MAX30100 measures light as a number from 0 to 65 535. So your pulse arrives as “46 012, 45 980, …”, not as a heartbeat yet. The next steps turn those numbers into meaning.Your code appears here after your first request.
Two ways to do it. Pick the one that suits you; both end with OrbitCore sending numbers over USB.
The fastest route for a workshop. This installs MakeSense’s ready-made MAX30100 firmware straight from the browser, with nothing to install. It sends infrared and red light 50 times per second, flashes the board’s LED on every heartbeat, and later (step 8) carries your trained model.
I2C. The MAX30100 talks to OrbitCore over two wires, SDA (data) and SCL (clock). The board writes settings into the sensor’s registers (which lights, how bright, how fast) and reads the measurements back.
The sensor keeps time. It measures exactly 50 times per second on its own clock and stores up to 16 readings in a small memory (a FIFO). The sketch empties that memory in every loop. Evenly spaced readings matter: later steps measure the time between heartbeats.
Two lights. Each reading has an infrared and a red value. Blood absorbs the two colours differently, so the two signals carry slightly different information.
One line per reading. Each line holds the time and the light values, separated by commas. That simple format (CSV) is exactly what a dataset is: rows of measurements.
Next: step 2: Collect
Board flashed and wired? No board today? Either way, continue: the next step also offers simulated signals.