Your smartwatch and similar accessories already gather a pretty impressive array of health data. They can offer clues about your blood circulation, sleep patterns, and even detect some conditions before noticeable symptoms emerge.
But your watch can only do so much. Now, a team of scientists and engineers from the Korea Advanced Institute of Science and Technology (KAIST) has developed a device that might just make this continuous conversation a reality. They have created a soft, flexible “smart patch” that sticks to your skin like a temporary tattoo but acts like a sophisticated mobile laboratory. It analyzes the chemical content of your sweat in real time and uses artificial intelligence to translate that data into a meaningful profile of your metabolic health.
A Microscopic Marvel of Engineering
The device looks deceptively simple. It’s a transparent, flexible disc that adheres to the skin. But beneath its unassuming surface, it has quite a bit of nano-engineering and microfluidic technology.
The patch consists of two main layers: a dermal-contact layer that sticks to the skin and collects sweat, and a “plasmofluidic” channel layer above it that does the heavy lifting of analysis. The magic begins when you start to sweat.
The sweat is whisked away from the skin and guided into a network of microscopic channels. This approach is known as microfluidics. But it doesn’t use just one open channel. The engineers designed a system for “chronological sampling,” meaning it captures sweat at different points in time to create a sequence. This is important because it doesn’t just show one snapshot of your health, but rather how it changes over time.
To accomplish this, the team designed a series of tiny chambers, each guarded by a microscopic gate called a capillary bursting valve (CBV). It’s a bit like a series of small dams on a river. The river is the sweat that flows in. First, it fills the first reservoir. Then, once that chamber is full, the pressure builds up until it’s enough to “burst” the valve and reach into the next chamber. The next is protected by a slightly stronger dam, and so on. Each chamber fills up sequentially over time, creating a physical timeline of your sweat captured in different compartments. This design ensures that the device can differentiate between a sample from now versus one from an hour ago.
But the analysis itself is even more complex.
Nano Islands
The floor of each chamber is not smooth; it’s covered by small “nano islands”. These are small studs that enable a powerful analytical technique called Surface-Enhanced Raman Spectroscopy, or SERS.
The process works like a chemical barcode scanner: a low-power laser is shone onto the sweat sample. The molecules within the sweat vibrate when hit by the laser light, scattering it back in a unique pattern. Each molecule has its own distinct vibrational “fingerprint.” The silver nano-islands act like miniature antennas, amplifying this scattered signal by over ten million times, making it possible to detect even very low concentrations of metabolites.
An important advantage of this method is that it’s “label-free,” meaning it doesn’t require adding any chemicals or enzymes to tag the molecules of interest, a limitation that hampers many other wearable sensors. The patch simply reads the intrinsic signatures of the molecules that are already there.
Creating this system on a flexible, rubbery surface that can act as a patch wasn’t easy, either. To accomplish this, the KAIST team laid down an ultrathin film of a Teflon-like material and deposited a thin layer of silver on top. With gentle heating, the silver film autonomously “beads up” and generates the nano-islands without losing elasticity.
Training The System
As impressive as all this is, the achievement is only half the battle. The data is still not easy to interpret and make sense of.
Sweat is a complex chemical cocktail. It contains salts, proteins, and dozens of different metabolites all mixed together. The resulting SERS spectrum is therefore incredibly complex and noisy. It is a chaotic chorus of molecular vibrations where the signal from one molecule can be drowned out or distorted by another. Even with this clever setup, simply looking for the peak intensity of a single molecule, a common practice, is often unreliable in such an environment.
This is where they turned to machine learning to make sense of the data. The team took a broad and robust approach, anticipating the chemical chaos of the mixture in real sweat. They created 41 different artificial sweat “cocktails.” Each had a unique and physiologically relevant combination of lactate, uric acid, tyrosine, glucose, and other background molecules. They then trained the AI on these cocktails, “teaching” the algorithm to recognize the unique fingerprint of a target molecule even when it’s surrounded by interference.
The patch also proved to be remarkably robust. It maintained over 85% of its performance after 25 days in storage. It withstood 200 cycles of bending and twisting with no notable degradation, and it could even survive being repeatedly peeled off with tape, proving its readiness for the mechanical stresses of real-world wear.
Putting it to the test
With the technology proven in the lab, the big challenge was still performance on an actual human being. For this, the team recruited four healthy participants for an on-body evaluation. The experiment was designed to induce clear metabolic changes. On one day, the participants exercised after fasting. On another day, they performed the same exercise routine, but only after eating a purine-rich meal — sardines, to be exact. Purines are natural compounds that, when broken down by the body, produce uric acid.
Participants went through a warm-up on a treadmill followed by a more intense session on a climb mill. As they exercised, the patches quietly and sequentially collected sweat samples.
After the sessions, the patches were detached and analyzed. The results captured the physiological story of each participant. They showed reliable information regarding compounds like lactate (a well-known byproduct of anaerobic metabolism), uric acid, and tyrosine, which is linked to the digestion of protein. The patch didn’t just detect changes in these compounds but quantified them (researchers confirmed with conventional lab equipment).
In the end, researchers say, the patch successfully tracked metabolic markers related to both exercise (lactate) and diet (uric acid). Such a tool could one day help athletes fine-tune their training routines, empower individuals with gout to manage their diet by seeing the immediate effects of certain foods, or provide personalized nutritional feedback for anyone looking to optimize their health. The era of listening to your sweat is just beginning.
Journal Reference: Jaehun Jeon et al, All-flexible chronoepifluidic nanoplasmonic patch for label-free metabolite profiling in sweat, Nature Communications (2025). DOI: 10.1038/s41467-025-63510-2
