Over the past two years, the AI-powered, wearable-integrated health and wellness firm has rebuilt around objective, continuous data—instead of asking consumers what they think is going on inside their bodies, its platform extracts data directly from their wearables, scores it and adapts a user’s supplement stack every month as the data changes.
This aligns with shifts in the personalized nutrition sector, which Mathieu Kremeth, co-founder of Nutricode, calls “a climb up a ladder of sophistication” for both companies and consumers.
“A decade ago, personalized nutrition mostly meant generic supplementation with a vague sense that something was good for you, with no real goal attached,” he said. “From there, consumers moved to goal-based thinking: take this specifically for sleep or for energy.
“Then came the realization that if a supplement isn’t doing anything, you should change it. Where the most engaged consumers are now is the top of that ladder, which is they want proof. They’ve become genuinely skeptical, and rightly so. Anyone who’s data-literate knows that perceived improvement can easily be placebo.”
The shift from belief to evidence is the single biggest behavioral change Kremeth has seen, and that is why he believes measurement—not just recommendation—is becoming the real battleground.
On the technology side, the enabling change was the move from snapshots to streams.
Ten years ago, personalization was built on one-off inputs, such as a DNA kit or a single blood panel. The proliferation of wearables, and more recently continuous glucose monitors and at-home testing, turned that into a constant flow of objective data.
“That’s a fundamentally different starting point: You can finally track whether an intervention worked rather than just prescribing it and hoping. This shift towards active measurement, meaning real-time continuous monitoring rather than static testing, already made up majority of the market in 2025.”
The maturation of artificial intelligence (AI) in parallel has also made it possible to interpret significant volumes of data for each individual.
According to Kremeth, the global personalized nutrition market value was estimated at US$16 billion in 2025 and projected to grow to US$ 36 billion by 2031 at around 14% CAGR, with the AI-specific slice growing at an even faster rate (close to 18% CAGR) through the early 2030s.
“The sharp rise in preventive health spending after the pandemic and GLP-1 medications pushing metabolic health to the front of the mainstream conversation in a way it never was before are two macro forces accelerating everything,” he said.
The net effect is that personalized nutrition has evolved from a niche wellness add-on to something much closer to a core, data-driven health tool.
Where opportunities lie
Kremeth reckons that the next decade comes down to proof. While most companies can now offer “tailored supplement stacks”, the unsolved problem is showing consumers what they are taking is in fact working, continuously and honestly.
“So the shift will be from personalizing the recommendation to closing the loop around it: measure, adapt, and demonstrate the result over time,” he said. “Whoever earns trust on that front wins the category because trust is precisely what this industry has historically lacked.
“That reframes where innovation lies. As wearables become near-ubiquitous, simply having data will stop being a differentiator and become table stakes. The competitive edge will come from how rigorously you interpret that data, the quality of the evidence behind your recommendations and how transparently you can show change over time without overclaiming.”
At the same time, there is a “real opportunity” within the longevity boom.
Kremeth explained that many people are buying longevity products while their fundamentals (sleep, recovery, resting heart rate) are in poor shape. Yet, longevity is essentially about maintaining and extending a good health state.
“If you’re not in a good state yet, chasing it is premature,” he said. “The opportunity is to educate people on that sequencing: fix and strengthen the baseline first, then work on extending it.”
This is also the pillar behind Nutricode’s two-phase approach: 1. Get the core metrics to a strong baseline and 2. Move into a longevity-oriented stack.
“I would also expect the wider industry to move that way rather than continuing to sell longevity as a shortcut, especially as education around the main longevity markers keeps growing and becomes more mainstream,” Kremeth said.
Another key frontier would be integration—pulling together wearables, continuous glucose and biomarkers, and other signals into one coherent picture rather than a scatter of disconnected apps.
“Geographically, Asia-Pacific is the region I’d watch most closely,” Kremeth noted. “Analysts expect it to be among the fastest-growing markets over the coming decade. The mix of insanely high digital adoption, a strong health-optimization culture and rising disposable incomes makes it a natural home for data-driven, subscription-based personalized nutrition.”
How Nutricode’s system works
Nutricode’s platform is said to be less a single off-the-shelf tool than a system built in-house.
With both founders, Kremeth and Romain Del Friuli, coming from AI research backgrounds, the deliberate choice from the start was to own the full pipeline rather than bolt together third-party products.
The system consists of three main layers. The first is the data layer, which integrates directly with the wearables through their official APIs, providing a clean, continuous stream of objective data rather than relying on self-reported inputs.
The second is the analytics layer, where raw data becomes meaningful—the Metric Health Score is computed and normalized against the user’s demographic, and a correlation analysis is run across metrics to surface the statistically significant relationships that point to the right target.
The recommendation logic is centered on an ensemble-learning approach, which combines several models rather than leaning on a single algorithm, making the output more robust.
The third is the knowledge layer. Nutricode’s research database is established from manual review and classification of over 1,000 peer-reviewed papers by health goal, supplement, population, dose, study type, duration and strength of outcome, with an evidence-confidence scoring framework.
“The core idea is simple: identify the weakest metric from the wearable data, choose the most evidence-backed intervention for that metric, then measure whether the expected change actually happens in real life,” Kremeth explained.
This framework lets the system pick the highest-quality intervention for a given target at the dose the research supports. It also stops recommendations from being generic. Instead of ‘supplement X is good for runners’, it’s ‘your data suggests deep sleep, HRV or VO2 max as the priority.’”
Finally, the system closes the loop by tracking the same metrics over the same time window in which research suggests measurable changes should occur, then evaluating whether the expected improvements are observed.
Addressing data privacy
As data form the foundation of Nutricode’s entire platform, the company places a strong emphasis on data privacy.
“We treat privacy as a design constraint from the start, rather than a compliance box to tick afterwards,” Kremeth said.
The company’s approach is guided by several principles. First, users retain ownership of the data, which is collected solely to generate and refine recommendations.
“We don’t sell the data to third parties, and a user can ask for it to be deleted at any time,” Kremeth said.
“The connection to a wearable is made through the device maker’s official authorization flow, which means we never see or store anyone’s device login. Access is granted by the user and can be revoked by them just as easily. Data is encrypted in transit and at rest, and internal access is limited to what is strictly needed to run the analysis.”
Operating across Singapore, Australia, Hong Kong and Malaysia, Nutricode adheres to the data-protection regulations in each market—specifically, Singapore’s Personal Data Protection Act (PDPA), Australia’s Privacy Act and the Australian Privacy Principles (APPs), Hong Kong’s Personal Data (Privacy) Ordinance (PDPO) and Malaysia’s Personal Data Protection Act (PDPA).
“As we are dealing with health and biomarker data, which are sensitive personal data, we hold it to a higher bar of explicit consent and protection than ordinary personal information.
“Ultimately, in a category where we are asking people to trust us with something as personal as their health data, we did not want to treat privacy as just another feature. We fully leaned into it, making it a core part of our product and value proposition. Because if users don’t trust how their data is handled, nothing else really matters.”
