Biopratiq  ·  Science · Innovation · Equity

The Body Keeps a Record. We Build the Systems That Read It.

The sensor that captures the signal, the method that reads it, the model that turns it into a decision. Built so that accuracy never depends on who you are, or where you are.

Scroll

01  Why We Exist

Biology is measurable. For most people, most of it goes unmeasured.

Signal before symptom A trace of a biological measurement over time. A small deviation appears early and is ringed and labelled SIGNAL. A much later point, where the trace has climbed steeply, is marked with a dashed line labelled SYMPTOM. SIGNAL SYMPTOM
The deviation is measurable long before it is felt.

Almost everything that goes wrong in a body announces itself as a signal before it announces itself as a symptom. A change in the skin barrier. A shift in what an organ puts into the blood. The signal is nearly always there first.

Reading it takes three things that rarely exist in the same place. An instrument close enough to capture the signal. A method honest enough to interpret it. And a model that holds up on the person standing in front of it, whoever that is. Where all three exist, medicine is precise. Where they do not, people wait for the symptom.

Biopratiq builds all three. A tool that only performs in the best equipped rooms was never accurate. It was only tested somewhere easy.

02  What We Build

One signal path, three layers.

Most of the field builds one layer and buys the others. We build the whole path, because a diagnostic is only ever as honest as its weakest link, and the weakest link is almost never the model.

SenseHardware

Measure the body directly

Instruments that take the signal from the source rather than inferring it from an image someone happened to capture. A photograph encodes the camera, the lighting and the skin tone at least as strongly as it encodes the biology, and no amount of post processing removes a confound baked in at capture.

ReadMethod

Interpret without inheriting the gap

Analysis built so that accuracy does not depend on who the person is. Equity as a term in the loss function rather than an audit at the end. Methods, weights and validation sets published open access, because a claim about fairness that cannot be checked is not a claim.

DecideModel

Put the answer where the person is

Output that a non specialist can act on, with the model's own uncertainty attached, running on constrained compute and intermittent connectivity. The room where the decision gets made is usually not a teaching hospital, and that is the room we design for.

Each layer is only as good as the one beneath it. That is why we build all three.

03  How We Work

A programme has to earn its place.

  • 01 We can measure the thing directly, rather than infer it.
  • 02 The measurement changes a decision someone is already making.
  • 03 The people that decision affects most are the ones current tools serve worst.

Programmes are added deliberately and rarely

04  Programmes

What we are working on

One active  ·  One in evaluation

Programme 01

Equitable Dermatology

  • Active
  • Research open access
  • Product in beta
  • Sensor in development

Skin is the organ we can reach without going inside anyone, and it is the one dermatological AI reads worst for most of the people on the planet. The programme builds the whole path for it: the sensor on the skin, the method that reads across the full Fitzpatrick spectrum, and the product that puts the answer in the hands of someone who will never see a specialist.

Programme 02

In evaluation

Nothing about the subject, and no date. What we can tell you is the bar it has to clear, which is the same bar Programme 01 cleared.

  1. 01 Measurable directly, not inferred
  2. 02 Changes a decision already being made
  3. 03 Serves the people current tools serve worst

We name a programme when it clears the gate. Not before.

05  Work With Us

Three kinds of people we want to hear from.